Surface-enhanced Raman scattering (SERS) technology has the advantages of high sensitivity, fast detection speed, and real-time analysis and is widely used in medical, biological, food safety, environmental monitoring and other fields. Currently, the detection methods of SERS signals of sample molecules mainly include single-point and long-range detection. The repeatability of the single-point detection method is easily affected due to the uneven distribution of sample molecules and the limited detection range of the laser spot. In order to make up for the deficiency of single-point detection, the long-range detection of Raman signals based on optical waveguides and optical fiber has been studied extensively in recent years. This paper summarizes the research progress of long-range detection of SERS signals in recent years and analyzes current long-range detection methods challenges and future development trends of current long-range methods. Firstly, this paper introduces the basic principles of single-point detection and long-range detection. On this basis, the research progress of long-range detection of SERS signals based on optical fiber is introduced. The long-range detection methods of SERS signals based on optical fiber include hollow fiber and solid fiber. The long-range detection method of SERS signals based on hollow optical fiber uses hollow optical fiber as the composite channel for liquid transport and signal transmission, which has an effective detection distance of centimeter order and high sensitivity. However, the detection method is difficult to inject, and the molecules of the sample to be measured in the composite channel are not easy to clean thoroughly; The long-range detection method of SERS signals based on solid fiber usually uses physical or chemical means to process the inherent structure of the solid fiber, and the detection distance is generally in the order of micrometers to millimeters, which is relatively difficult to manufacture. Then, the research status of long-range detection of SERS signals based on optical waveguides is summarized. The long-range detection of SERS signals based on liquid-core optical waveguides combines microfluidics with SERS, which can effectively increase the contact area between sample molecules and SERS “hot spots” and improve its detection sensitivity. This method can reach the level of single-molecule detection, but there are difficulties in preparing enhanced media in microchannels. Most of the long-range detection of SERS signals based on solid-state optical waveguides is currently in the theoretical analysis stage, and the long-range detection structure of SERS is often studied and analyzed through simulation software to explore its process mechanism. Finally, the research progress on the long-range detection of SERS signals is summarized and prospected, and feasible research suggestions are put forward to provide a reference for the related research on the long-range detection of SERS signals.

the progress of society, peoples dietary requirements are constantly improving, which is gradually changed from the previous “eat full” to todays “eat well”. People are paying more attention to food safety. Therefore, fast and non-destructive food detection technology is needed to meet the imminent demand for food safety. Spectral technology can calculate the material characteristics and composition of food samples according to their physical structure and chemical composition. It has a broad application prospect in adulteration detection, freshness detection, and residue detection of harmful substances. Compared with the traditional detection technology in food detection, spectral technology has the advantages of rapid, high precision, no sample loss, and good repeatability, and it has become an important development direction in food detection. In this paper, related domestic and international literature on spectral techniques applied to food detection in the last five years is discussed, focusing on data pretreatment method, characteristic band selection algorithm and data modeling method to systematically review the application and progress of spectral technology in food detection. In this paper, the application of spectral technology in food detection is discussed, including the preprocessing of spectral data by multiplicative scatter correction (MSC), standard normal variate transform (SNV), and Savitzky-Golay smoothing (SG) algorithm; successive projections algorithm (SPA), principal component analysis (PCA), and competitive adaptive reweighted sampling (CARS) were used to select characteristic bands; partial least squares (PLS), support vector machine (SVM), and artificial neural network (ANN) were used to analyze collected data. Simultaneously, this paper summarizes the prospects for the application of spectral technology in food detection: the integration of spectral detection technology and a variety of food detection technology will become a new development direction in the future; combining spectral detection technology with on-line detection technology to realize on-line and real-time detection of food samples will obtain more valuable detection results; the development of portable spectral detection equipment will be more convenient for on-site food detection, and this equipment will significantly improve the efficiency of food detection and has excellent market potential.

Bioaerosols are aerosols containing biological particles such as bacteria, fungi, pollen, etc. The spread of bioaerosols has a potential impact on human health and the atmospheric environment. In addition, bioaerosols are also used as the release mode of biological agents in military activities. Therefore, real-time detection of bioaerosols in the air, rapid identification of aerosols types, and determination of concentrations and dangerous levels of bioaerosols are important methods to reduce exposure to pathogenic bioaerosols, protect personnel and environmental safety, and prevent bio-terrorist attacks. Biological particles contain typical fluorophores such as tryptophan, tyrosine or riboflavin. The feature fluorescence spectrum can be obtained via laser-induced these biological substances, thus completing the detection and identification of biological aerosols. The real-time detection system of bioaerosols based on the fluorescence method offers significant technological advantages in identifying aerosols biological and physical characteristics. The basic principles of measuring bioaerosols are briefly introduced. The research on the real-time detection system of bioaerosols is summarized in three aspects: the triggering method of the fluorescence excitation light source, the type of the fluorescence excitation light source and the signal acquisition system. Finally, the development direction of the real-time detection system of bioaerosols is discussed, which provides a reference for the subsequent research of the real-time detection system of bioaerosols.

Biological detection is a common method in biomedicine based on the change of specific biological activity of cytokines. It is mainly applied in biomedicine, agriculture and forestry, etc., and plays an important role in the study of medical pathology and the law of crop diseases. Traditional detection methods mainly make judgments by observing the reaction of test samples to different chemical reagents. Although the detection accuracy is high, problems include cumbersome operation and long detection cycle. Hyperspectral imaging technology combines optical imaging and spectral analysis, which can simultaneously obtain the image data and spectral information of the detected samples. The image data in hyperspectral images reflect the samples external characteristics and surface texture, while the spectral information can analyze detected samples the internal physical structure and chemical composition. Micro-hyperspectral imaging technology is a fast, nondestructive and accurate optical imaging analysis technology that combines hyperspectral imaging technology with the biological microscope to analyze the detected samples by observing the microscopic worlds image data and spectral information. In recent years, because of its high resolution and continuous data, microscopic hyperspectral imaging has attracted more and more attention in biological detection and has become one of the important means of biological detection. Based on the basic principle of spectral imaging, data processing and application of biological detection, this paper reviews the research status of microscopic hyperspectral imaging technology in biological detection in recent ten years. It puts forward some problems existing in the research process of microscopic hyperspectral imaging technology based on summarizing the research achievements. The future development trend of microhyperspectral imaging in biological detection prospects to provide a reference for the research and application of microhyperspectral imaging in biological detection.

With the rapid development of modern remote sensing techniques, remote sensing image change detection has become one of the most important means of the land-cover monitoring process. It has been widely used in application areas such as Geographic Situation Detection, Land Survey, Ecosystem Monitoring, Disaster Monitoring and Assessment, Food Security Insurance and military reconnaissance. The fine spectral resolution of the hyperspectral (HS) image and the detailed spectral change information of the multitemporal HS images brings the possibility for detecting the subtle changes associated with the dynamic land-cover transition. However, the high complexity data structure, high dimensional data features, and high redundancy information of the HS images makes HS change detection extremely challenging. This paper reviews the research advance of multitemporal HS image change detection, including: (1) Traditional HS image change detection approach based on the generalized similarity measurement of the HS images, which mainly follows the modeling process of multispectral change detection methods; (2) Dimensionality reduction based HS image change detection approaches, which are designed to overcome the adverse effects of the high dimensionality, high redundancy properties of the HS images; (3) Statistical modeling based HS image change detection approaches, which determines the change detection results by modeling of the statistical properties and multi-dimensional correlations of the HS images; (4) Classification based HS image change detection approaches, which introduces the image classification strategy into the change detection process to provide guarantee for obtaining the “from-to” type change information; (5) Unmixing based HS image change detection approaches, which are mainly developed to solve the mixed pixel phenomenon caused by the low spatial resolution of HS images; (6) Deep learning based HS image change detection approaches, which applied the deep learning methods into the HS image change detection tasks. Finally, the three major challenges and future development of HS image change detection are prospected.

Quantum cascade laser (QCL) plays an important role in mid-infrared detection because of the high output power and wide coverage range. However, due to the fluctuation of laser wavelength caused by the sensitivity of the laser to changes in the external environment, the peak-to-peak frequency drift is as high as 180 MHz within the observed time of 400 s, which affects the performance of the QCL to some extent and reduces the accuracy of molecular spectral detection. Frequency locking has been widely applied to the mid-infrared areas. In this paper, a QCL frequency locking system based on gas absorption was developed. Taking 5.3 μm QCL as an example, the laser frequency is locked to the absorption peak of nitric oxide (NO) molecule at 1 875.812 8 cm-1 by modulating laser wavelength. The principle of error signal generation was introduced, and the advantages of using the third harmonics as an error signal for frequency locking were analyzed. The NO absorption signal with a high signal-to-noise ratio (SNR) was obtained using a NO absorption cell with a length of 30 cm. The conversion coefficient between the third harmonic voltage and the laser frequency was calibrated. The locking process was introduced in detail and explored the significance of proportional, integral, differential parameters of the feedback loop during the locking process, and the locking parameters had been given in detail. Disturbing the locking system, with the recovery time better than 40 ms demonstrate that the locking system can respond quickly and remain stable against external disturbances. In addition, the stability of the frequency locking system was also verified by the fluctuation of the error signal with the voltage-frequency conversion coefficient. A frequency drift better than 673 kHz (1σ, 10 ms integration time) was achieved. The Allan variance analysis results show that when the integrated time is extended to 100 s, the frequency drift is lower than 4.5 kHz (corresponding to stability of 8×10-11), effectively improving the laser frequencys long-term stability. This method of directly modulating laser frequency without an external modulator simplified the system and improved the stability of the optical detection system.

In the photovoltaic field, LED has been proven to be a promising artificial light source, and its spectrum can synthesize sunlight spectrum. At present, the key problem of synthesizing the sunlight spectrum through different kinds of LEDs is how to select led models from the LED database and determine the working parameters of each model of LED. To solve this problem, scientists have proposed various mathematical spectral models to express the spectral power distribution of LED, such as the Gaussian function, Lorentz function, logistic power peak function and so on. These mathematical models can accurately reflect the nonlinear spectral characteristics of LED, including peak wavelength, FWHM, redshift and so on. However, the LED spectrum simulated by the mathematical spectrum model is different from the real led spectrum. The LED spectrum curve simulated by the mathematical spectrum model is symmetrical. However, the spectral curves of real blue and green LEDs are steeper on the short wavelength side and slower on the long wavelength side, while the spectral curves of red LEDs are the opposite. Based on the existing LED light sources, this paper proposes a new method to simulate the solar spectrum. The simulation experiment is carried out by designing a filter and adjusting the radiation weight of each existing LED light source (the radiation weight of an LED light source is the percentage of the radiation of the current LED light source and the maximum radiation of the LED), to simulate a new light source and make its spectrum close to the sunlight spectrum. That is, one filter is placed under the mixed light source of multiple LEDs, and the spectrum of the light source is limited by the spectral transmittance of the filter and the radiation weight of each LED light source so that the spectrum of the modified multiple LED mixed light sources is as close to the spectrum of the daylight light source as possible. Due to industrial requirements, it is necessary to smooth and constrain the spectral transmittance curve of the filter. Finally, through the spectral transmittance of the optimized filter, the least square solution of the radiation weight of each LED light source is obtained to obtain the radiation weight of each optimal LED light source. Through the simulation experiment, the correlated color temperature (CCT) of the simulated sunlight light source obtained by this method is 6 492, and the correlated color temperature of the target light source is 6 503. The correlation index (R2) between the simulated sunlight source and the sunlight spectrum is 0.992 6. The method of this paper provides an important reference value for the research of solar spectrum fitting based on LED.

The photoacoustic cell is where the “light heat sound” coupling occurs. The performance of photoacoustic cells directly affects the accuracy and sensitivity of the detection system. In order to improve the performance of photoacoustic cells, a two-stage buffer photoacoustic cell is proposed based on the traditional cylindrical photoacoustic cell. Through the simulation of the thermal viscous acoustic physical field interface in COMSOL software, the effects of the height and number of buffer partitions on the sound field in the photoacoustic cell are analyzed.The results show that the resonance frequency of the photoacoustic cell decreases with the increase in the number and height of buffer partitions. When the number of buffer partitions is fixed, the height of the buffer partition is greater than 11mm, and the resonance frequency of the photoacoustic cell decreases sharply with the increase of the height of the buffer partition. In the required range of photoacoustic cell resonance frequency, the decrease of the resonance frequency is conducive to the increase of photoacoustic signal amplitude; When the height of the buffer partition is fixed, the sound pressure of the photoacoustic cell decreases with the increase of the number of buffer partition; When the height of the buffer partition is between 0 and 11mm, the sound pressure value remains relatively stable; When the height of the buffer partition is greater than 11mm, the sound pressure decreases sharply with the increase of the height of the buffer partition. Regarding flow field, the velocity gradient in the left buffer cavity can be reduced by setting a buffer partition in the cavity. Although the one-stage buffer can reduce the velocity gradient to a certain extent, there is velocity fluctuation at the buffer partition. The two-stage buffer reduces the velocity gradient in the photoacoustic cell and makes the gas flow more stable. Considering the photoacoustic signal amplitude, sound pressure and velocity gradient in the photoacoustic cell, the height of the buffer partition is 11mm and the number of buffer partitions is 2.Based on the optimal parameters given, the simulation and experimental results show that the sound pressure of the two-stage buffer photoacoustic cell is 3.34×10-5 compared with the cylindrical photoacoustic cell of the same size reduced to 3.32×10-5, background noise from (2.83±0.11) μV decreases to (1.26±0.03) μV. The resonance frequency is reduced from 1 344 to 1 299 Hz. Although the sound pressure is reduced by 1.2%, the sound-to-noise ratio is increased by 2.22 times, and the resonance frequency is reduced by 3.3% within the range meeting the requirements so that the amplitude of the photoacoustic signal is improved to a certain extent. Overall, the two-stage buffer photoacoustic cell stabilizes the gas flow noise and reduces the fluctuation range of the flow noise. The proposed two-stage buffered photoacoustic pool provides a new idea for the optimal design of a photoacoustic pool.

As grassland degradation and desertification are becoming more serious in the northern Tibetan plateau, estimating gravel grain size is important for desertification evaluation and dynamic monitoring. Based on hyperspectral remote sensing technology, this paper combines ground survey, GPS positioning, gravel spectroscopy and gravel grain size determination, preferably selecting the waveband with the highest correlation with gravel grain size, and establishes a linear fitting model between gravel grain size and sensitive waveband, and extracts the spatial distribution characteristics of gravel grain size in the experimental area using hyperspectral images of HMS-5. The results show that: the bands with better correlation are at 369.9, 371.5 and 910.5 nm, where the first-order derivative at 910.5 nm has the best fitting effect with gravel grain diameter (R2=0.738); the fitting comparison of different spectral absorption parameters with gravel grain diameter, the fitting accuracy of the fitted regression model established by the absorption area near 2 340 nm and the gravel grain diameter is The fitting accuracy of the fitted regression model is relatively high (R2=0.728); in the inversion of gravel grain size by spatial remote sensing, the accuracy reaches 70%, as well as briefly analyzing the spatial distribution characteristics of gravel grain size, which provides a reference basis for the desertification analysis of the northern Tibetan plateau.

Rice mold can cause nutrient loss and produce toxic substances that reduce its quality and infect other normal rice. In order to reduce the loss of rice caused by mold, moldy rice needs to be separated promptly. Hyperspectral technology is fast and nondestructive, so an attempt was made to detect rice mold using hyperspectral technology. Germinated rice and moldy rice have similar spectral characteristics and are easily misidentified as moldy rice, which affects the subsequent detection of rice mold degree. Therefore, it is proposed to use hyperspectral techniques combined with various pre-processing and discrimination models to distinguish germinated rice from moldy rice and to discriminate rice with different mold degrees. Sound, sprouted, moldy and germinated moldy rice samples were modeled to differentiate and detect mild, moderate, heavy and completely moldy rice samples. The spectral images of sound, germinated, moldy and mildewed rice samples were acquired using a hyperspectral acquisition instrument to extract the spectra in the region of interest (ROI) of the acquired images, and the average reflectance of the spectra within the ROI was used as the spectral characteristics of the rice samples. Pretreatment of the extracted spectral data with SNV, Normalize and MSC. The KS algorithm is used to divide the samples evenly in a ratio of 1∶3, into a prediction set for validating the effect of the model and a modeling set for establishing the relationship between the spectra and the samples. The PLSR, SVM and RF models were developed respectively, and the prediction effect of each model was evaluated by the prediction set correctness of the three models, and the discriminative model with the best effect was selected. In detecting sound, germinated, moldy and germinated moldy rice, the optimal discriminatory model was obtained as a random forest (Baseline-RF) model after pre-treatment by the baseline correction method. The discriminatory accuracy of the prediction set of the Baseline-RF model was 100%. In detecting rice mold degree, a comparison of the prediction results of different models showed that the SNV-RF model showed the optimal discriminative effect with no misclassified samples in the prediction set. The characteristic wavelengths were extracted from the lengthy original spectra to simplify the model, and the SNV-RF model was established with the spectra under the characteristic wavelengths. The results showed that the characteristic wavelengths selected using the CARS algorithm had good discriminative ability, and the overall discriminative accuracy was 97.5%. The experimental results show that the hyperspectral technique combined with the CARS-SNV-RF model can quickly and accurately discriminate the degree of moldy rice, which provides a certain theoretical basis and experimental reference for the rapid discrimination of moldy rice and is of great significance for improving the quality of rice and reducing the waste of rice.

The trade of elephant ivory (from now on referred to as ivory) and its products has been completely banned in China, but the smuggling has not stopped. The identification of ivory is an important part of the fight against smuggling. The ivory objects can be identified visually if characteristic features, such as the Schreger pattern, on the cross-sections can be observed. However, if it does not have the features or disappears after carving, the sample is difficult to identify by morphology. When the morphological method is not adequate, molecular biology-based method is an alternative. However, since the extremely low DNA content, extracting DNA from ivory is not easy. Some scholars have distinguished ivory and the analogues (teeth of other animals) by using the Raman spectrum and the short wave region（780~1 100 nm）of the near-infrared (NIR) spectrum. In this study, an identification method of ivory was established by establishing a NIR spectroscopy identification model, based on the spectrum region from 1 000 to 1 800 nm. Taking African elephant (Loxodonta spp.) ivory, Asian elephant (Elephas maximus) ivory and mammoth (Mammuthus spp.) ivory as the calibration set and other non-ivory products, including sperm whale (Physeter macrocephalus) teeth, hippopotamus (Hippopotamus amphibius) teeth, taguanut (phyelephas macrocarpa), ivory plastic imitation, etc. as the verification set, a total of 383 NIR spectra of 230 samples were collected. By comparing the spectral data of ivory products with different colors and thicknesses, it was found that the color and thickness of the ivory will affect the analysis results. Scanning the parts with typical color of the substance and the parts with the thickness greater than 1 mm is recommended. Based on the spectra in the regions of 1 160~1 200, 1 430~1 500, 1 680~1 710 and 1 720~1 750 nm, and the SIMCA (soft independent modeling by class analogy) qualitative analysis method, a calibration model for predicting ivory or non-ivory by NIR spectroscopy was developed. After balancing the false positive rate and false negative rate, it was deduced that the best principal factor of this model was 2 and the F value was 0.21. When applying the model, the recognition accuracy of ivory was 100%; all the animal horn, plastic and ivory fruit products could be accurately identified as non-ivory with 100% accuracy. However, the teeth of other animals with similar ivory textures, such as boar teeth and sperm whale teeth, tended to be mistaken for ivory by the model; consequently, further testing by other methods was required for these substances. The spectral model method is simple, nondestructive, and more objective and efficient than manual interpretation of spectrograms. Therefore, it is suitable to be used as a preliminary screening method for on-site law enforcement by regulators.

Breast cancer is a very dangerous disease for women worldwide, its prevalence is increasing year by year, and it is the main cause of death among women worldwide. In the case of large samples, the clinical diagnosis of breast cancer is limited by the relative shortage of high-quality medical resources, the diagnosis cycle is long, and the detection cost is high. Therefore, efficient, accurate and cost-effective breast cancer diagnosis methods have broad application prospects and are urgently needed for clinical diagnosis. Fluorescence spectroscopy is a method that can characterize the combined physical and chemical changes in cells and can be used to characterize normal and cancerous cells. Machine learning is good at mining useful information from a large amount of data and is an effective classification and prediction method. In the past, machine learning mostly used databases containing some biochemical information to train models, which easily led to information loss. The fluorescence spectrum is the superimposed spectrum of multiple substances in cells, and the use of fluorescence spectrum characteristic peaks to diagnose breast cancer has the problem of quantitative uncertainty.Therefore, this paper proposes a diagnostic method combining machine learning with fluorescence spectra of breast cancer samples. The fluorescence spectrum data of normal and cancerous breast tissue (pathological diagnosis has been made) was collected as a data set, and K-nearest Neighbor (KNN), support vector machine (SVM), Random Forest (RF) three algorithms to classify the fluorescence spectrum of normal and cancerous breast tissue. The discriminant results show that compared with the SVM algorithm, the KNN and RF algorithms have higher accuracy, stronger ability to balance recall and precision, and better classification ability for breast cancer fluorescence spectra. The results of the F1-score function are all above 95%, which is more conducive to the diagnosis of breast cancer. Furthermore, the classification ability of the Weighted K-nearest Neighbor (WKNN) algorithm for normal and cancerous breast tissue fluorescence spectra were discussed. Compared with the KNN algorithm, WKNN has a small improvement in the classification evaluation results and has better anti-noise and adaptability, and the algorithm is simple. In conclusion, the breast cancer diagnosis method based on machine learning and fluorescence spectroscopy proposed in this paper has high accuracy, high speed and high-cost performance. It is the future development direction of breast cancer diagnosis methods and has important clinical application value.

The multi-axis differential absorption spectrometer (MAX-DOAS) combined with computed tomography (CT) reconstruction algorithm can be used to obtain the spatial distribution of the target trace gases. In order to study the feasibility of reconstructing the spatial distribution of NO2 on a vertical cross-section under the condition of background concentration (such as the urban background), a confirmatory experiment was designed under the condition of controllable gas density. The feasibility of using MAX-DOAS to reconstruct the distribution of NO2 in the vertical plane is proved. The JGS1 quartz glass sample cell filled with standard gas was used as the research object, and two MAX-DOAS were used to collect spectral data. Taking the gradient of gas density as a priori information, and using the classical ABOCS algorithm and the Barzilai-Borwein algorithm, the NO2 distribution in the vertical plane is reconstructed. The feasibility of using MAX-DOAS to reconstruct the spatial distribution of NO2 in the vertical plane was verified, and the influence of background density on the reconstruction results was determined. The results show that the NO2 density obtained by the retrieving is very close to that by using the sky as the reference spectrum and the empty sample pool as the reference spectrum. Therefore, the influence of the sample pool container on the experimental results can be neglected in the reconstruction method of NO2 vertical plane distribution. The background density of MAX-DOAS with the urban background was high, especially when the observation angle was low, the background density of NO2 was almost 6×1016 molec·cm-2, the MAX-DOAS, whose background density has no obvious pollution source in the city suburb, is low enough to be ignored. The reconstruction results show that the average molecular number density along the optical path is 3.932 7×1015 molec·cm-2 when the observation angle is 28°, and the density in the lower region of the sample pool is higher than that in the upper region. The reconstructed SCD is in good agreement with the measured SCD. The calculated results show that the peak value of molecular number density is 5.77×1015 molec·cm-2, which is close to the MAX-DOAS retrieving results with the background of the urban suburbs. However, it is not close to the MAX-DOAS retrieving results with the background of the urban areas, especially when the elevation angle is small, whose background density of NO2 is especially obvious, and its reconstructed result is much lower than the measured one. To sum up, background density is an artifact in the reconstructed image, which affects the observation of gas distribution. If the prior information of gas density mutation in and out of the sample pool is added to the reconstruction algorithm, the effect of background density on the reconstruction results can be reduced.

The fibre content of sugarcane is a non-negligible factor in the breeding process of sugarcane and the production of sugar, paper and other industries. It is of great significance to detect the fibre content of living sugarcane non-destructively by using visible-near infrared spectroscopy in transmission form. One hundred and twenty-three sugarcane samples of six varieties at different growth stages were collected and divided into a calibration set (82 samples) and prediction set (41 samples) in the ratio of 2∶1 using the Duplex sample set division method. The transmission spectra of sugarcane cane stem in the original state and dewaxing state were acquired at a measurement angle of 120°, and the band from 670 to 950 nm with less noise and obvious amplitude fluctuations was chosen as the actual modeling band. The waveforms were observed to find a significant increase in transmittance after dewaxing, and a PLS (Partial least squares) regression model was established to analyze the effect of wax coverage on the predictive model ability. The sugarcane samples were modeled more effectively after dewaxing. The 9 preprocessing methods, including first derivation (FD), continuous wavelet transform (CWT), and standard normal transform (SNV), are divided into four steps: baseline correction, scattering correction, smoothing, and scale scaling. The order of the steps was permuted to produce 108 combined preprocessing methods, and the PLS modeling analysis was performed for the spectra after each combined preprocessing separately, finally, the preprocessing method FD+SG with the best comprehensive modeling effect was acquired. To screen for the wavelengths carrying the most effective information, effective variables screening algorithms such as uninformative variable elimination (UVE), genetic algorithm (GA), competitive adaptive reweighted sampling (CARS), and random frog algorithm (RF) were taken to select important wavelengths for the transmission spectra after optimal pretreatment. The important wavelengths extracted by each algorithm were analyzed for PLS modeling separately, in which the important wavelengths extracted by the UVE method were modeled the best, with the number of selected wavelengths being 40, accounting for 14.3% of the full band. The R2p is 0.73, which improves 14.1% over the full-band modeling results with the same preprocessing method, and the RMSEP is 0.88, which decreases 14.6% over the full-band modeling results. The results showed that the transmittance visible-NIR spectrum could effectively predict the fibre content of living sugarcane. This study can provide a theoretical basis for developing corresponding portable sensors, and provide technical support for sugarcane breeding and production efficiency in various industries.

The lipophilic azo dye Sudan Red Ⅲ molecule enhances flavor or makes food bright and attractive. After eating, it has noticeable toxic effects on the liver and kidney organs of the human body and seriously affects human health. The toxicity of Sudan red molecule is closely related to its molecular geometry and electronic structure, which has important guiding significance for studying its structure and electronic excitation. In this work, we aimed to systematically investigate the molecular structure, infrared and Raman spectra, and ultraviolet spectra of Sudan Red Ⅲ by using the density functional theory (DFT) method in conjunction with the def2-TZVP basis set. The excitation properties of Sudan Red Ⅲ were also studied in detail by the hole-electron analysis method. The results show that the infrared and Raman spectra calculated for Sudan Red Ⅲ using the PBE0 and B3LYP exchange-correlation functional agree with the experimental data. Using the time-dependent B3LYP (TD-B3LYP) method with the def2-TZVP basis set, The UV-visible absorption peaks of Sudan Red Ⅲ show 228, 353, and 490 nm, which are in good agreement with the experiments. It can be found that they are through the transition of the ground state electrons to the second excited state, the sixth excited state, and the 30th excited state. The electron excitation characteristics are studied by using hole-electron analysis. It can be found that S0→S2 is attributed to the superposition of the n—π* charge-transfer excitation in the direction from oxygen and nitrogen atoms to the naphthalene and benzene ring, and the π—π* local excitation between intra-ring naphthalene and benzene rings. The superposition of the n—π* charge-transfer excitation from oxygen and nitrogen atoms to naphthalene and benzene ring and the π—π* local excitation between intra-ring naphthalene and benzene rings are excited by S0→S6. The electronic transition of S0→S2 and S0→S6 from the ground state to the excited state belongs charge transfer excitation, where the charge transfer excitation effect is dominant. S0→S30 is attributed to the superposition of local excitation and charge-transfer excitation, where the local excitation effect is dominant. They contributed π—π* local excitation between intra-ring naphthalene rings, and the superposition of n—π* in the direction from oxygen and nitrogen atoms to naphthalene and benzene ring and π—π* charge-transfer excitation between intra-ring naphthalene rings. Furthermore, we draw heatmaps of the contribution of molecular fragments to holes and electrons. The electron mentioned above excitation transfer was further confirmed by heat map analysis. The spectrum and electron excitation of Sudan Red Ⅲ were investigated systematically, which can provide a theoretical reference for experimental detection of Sudan Red Ⅲ in food.

As a new laser surface treatment technology, laser shock peening (LSP) has been applied to strengthen and prolong the life of critical components such as the aero-engine and gearbox. Ensuring the consistency and stability of LSP processing quality is of great significance to the long service life of the aviation mentioned above equipment. However, the protective coating is easily ablated and damaged in the high-energy transient LSP process, limiting the industrial application of LSP significantly. Therefore, this paper proposed a real-time detection method for protective coating damage of LSP based on ReliefF feature weight fusion by analyzing laser-induced plasma spectrum signal. The 7075 aluminum alloy with a thickness of 4 mm was used as an LSP target, and the black tape was used as the protective coating. Firstly, the Ocean Optics-HR4000 spectrometer with a wide wavelength range and Princeton SP2750 spectrometer with a high resolution was used to synchronously collect the plasma spectra produced in the LSP transient process.Secondly, according to the spectral signals with a high resolution collected by the Princeton SP2750 spectrometer,the peak intensities and Stark-FWHMs of Al Ⅰ spectra at 394.40 and 396.15 nm and that of Fe Ⅰ spectra at 393.36 and 396.80 nm were extracted, then, combined with the feature importance obtained by the ReliefF algorithm to screen out two Al Ⅰ spectra that were more sensitive to the damage states of the protective coating, andqualitatively analyzedthe sensitivity and the law of transient variation for intensity and Stark-FWHM of Al Ⅰ spectra to the damage states of the protective coating. Furthermore, a feature named intensity-FWHM (I-FWHM) fusing the information of multiple spectral emission lines was constructed using the ReliefF algorithm, and then, the ability of each feature to distinguish the three kinds of damage states was quantitatively evaluated based on the distance between classes. Finally, combined with the threshold segmentation method, real-time detection of LSP protection layer damage was realized. The experimental results show that the intensity is very poor in distinguishing between normal state and slight damage but very strong in distinguishing between slight damage and complete damage. The Stark-FWHM is far better than intensity in distinguishing between the normal state and slight damage, but the abilityis relatively weak in distinguishing between slight and complete damage. The I-FWHM combines the advantages of the above single feature and can better distinguish the three types of damage states simultaneously. Therefore, I-FWHM has stronger anti-interference ability and higher robustness for the real-time detection of protective coating damage state in the LSP process.

Eating bird′s nest (EBN)can regulate intestinal flora and improve human immunity. Excavating the key nutrients and effective functional groups in fresh stewed bird′s nest (FSBN)can lay a research foundation for exploring the mechanism of bird′s nest regulating intestinal flora. The functional groups and material composition of the two brands of FSBN products were analyzed by Fourier Transform Infrared Spectroscopy (FTIR) and high-resolution mass spectrometry (HRMS), respectively. In order to explore the effective functional groups and key regulatory substances in the regulation of intestinal flora after eating FSBN, the relationship between FTIR and HRMS results and the abundance of intestinal flora in C57BL/6N mice was established by correlation analysis. The results showed that after taking FSBN, the intestinal bacteria that differ greatly in the species level are as follows: beneficial bacteria Lactobacillus and Lachnospiraceae_NK4A136_group showed an upward trend, beneficial bacteria Faecalibaculumand Akkermansia showed a downward trend, and harmful bacteria Desulfovibrio, Enterorhabdus and Candidatus_Saccharimonas showed a downward trend. The infrared difference spectra of the two products were mainly concentrated in the 1 700～1 200 cm-1 band, which were mainly the distribution areas of the amide bond, aromatic CC and carboxylic acid CO group. These functional groups were positively correlated with beneficial bacteria Lactobacillus and Lachnospiraceae_NK4A136_group and negatively correlated with harmful bacteria Desulfovibrio and Enterorhabdus. After analysing small molecular substances in FSBN by metabolomics, it was found that the main components were amino acids, lipids, esters and sialic acids, which showed different correlations with intestinal bacteria. Generally speaking, lipids, especially some phosphate esters (Phosphate, Phosphatidylcholine, Phosphatidylethanolamine and Phosphatidylglycerol) and fatty acyl showed extremely significant negative correlation with harmful bacteria and a positive correlation with beneficial bacteria. The characteristic groups in these substances can also correspond to the FTIR results. The above results show that both Fourier transform infrared spectroscopy and metabolomics can provide a theoretical basis for studying nutrients on the regulation of intestinal flora.

Hyperspectral detection is an important method for qualitatively identifying substances, and spectral unmixing is the key to hyperspectral analysis and identification. The blind source unmixing separation method based on weighted non-negative matrix factorization (NMF) hyperspectral reflection curves are established for the spectral decomposition and identification of minerals after mixing by using the NMF blind source unmixing method to address the problem of inaccurate analysis of compound or mineral mixed spectral in the paper. The algorithm assumes that the spectral mixing model is a linear combination of scaled component spectral signals, uses the minimum Euclidean distance and reweighted sparsity constraints to establish the combination conditions to promote the sparsity of the unmixing matrix, and carries out the iterative calculation of the unmixing NMF constraint with the initial weight of the spectral angular cosine of the mixing spectra and component spectral basis vectors to finally decompose the source spectral basis vectors and the abundance matrix of the mineral mixing spectral.Three mixtures of chemically pure CuO and Cu2O,Cu(OH)2 and Cu2(OH)2CO3, malachite and azurite hyperspectral profiles were selected for spectral unmixing and identification experiments. After the measured mixture spectral curves were equalized and whitened, the blind source unmixing calculation based on the weighted NMF hyperspectral reflection curve was carried out, and the unmixing performance index PI, the root mean square error of the spectral and the angular distance of the spectral were selected as the evaluation indexes of the unmixing effect. The experimental results show that the blind source unmixing effect of the NMF unmixing method is very obvious, the base source spectral features can be accurately separated based on unknown mixed spectral a priori conditions. The sample separation accuracy is less than 0.15. The curves of the de-mixed and the source spectral have the same overall trend, maintaining similar absorption positions and absorption peaks of the source spectral, with minor shifts and obvious differences in reflectance values of the corresponding absorption positions. After adding 5%~15% Gaussian noise to the mixed spectral data, a weighted NMF-based unmixing process was performed, and it was found that the unmixing separation accuracy decreased slightly as the noise increased. However, the overall angular distance of the spectral and the root mean square error did not change significantly after unmixing, indicating that the NMF unmixing algorithm has good noise immunity and applies to spectra unmixing of measured non-pure material, which provides a basic theory for the identification and separation of mineral components after mixing.

With the increasing demand for oil resources by economic development, oil pollution problems have become increasingly serious, posing a huge threat to the ecological environment and human health. Therefore, accurate identification and timely treatment of oil pollutants is significant in reducing oil spill hazards. Petroleum is a complex organic compound mainly composed of aromatic hydrocarbons and their derivatives with strong fluorescence characteristics. Different types of petroleum contain different components and contents of polycyclic aromatic hydrocarbons. Three-dimensional fluorescence spectroscopy 3D-EEM is widely used to detect petroleum pollutants. Based on three-dimensional fluorescence spectroscopy, the improved wavelet threshold function and BP(backpropagation) neural network combined with the method of self-weighted alternating trilinear decomposition (SWATLAD) algorithm for qualitative and quantitative research on oil pollutants. The experiment used 0# diesel, 95# gasoline and kerosene as the research objects. Firstly, the samples were detected using an F-7000 fluorescence spectrometer, and the obtained data were processed by excitation, and emission correction. Secondly, an improved threshold function is proposed to solve the problem of signal discontinuity and excessive shrinkage of wavelet coefficients at the threshold of wavelet threshold denoising. The signal-to-noise ratio (SNR) and mean square error (MSE) are 18.354 7 and 10.261 7, respectively, which can more accurately restore useful signals. The preprocessed spectral data were trained by BP neural network based on error backpropagation. After training, the curve of the predicted value after training was in good agreement with the real value, indicating that the subsequent fluorescence data collected by the spectrometer can be directly input into the neural network to output the preprocessed data, simplifying the experimental operation steps. Finally, Finally, SWATLD was used to decompose the data processed by improved wavelet transform and BP neural network. The excitation and emission spectra of 0# diesel, 95# gasoline and kerosene obtained by the analysis were in good agreement with the real spectra, and the calculated average recoveries were 103.64%, 99.33% and 97.85%. It is proved that three-dimensional fluorescence spectroscopy combined with improved wavelet transform and BP neural network can detect fluorescent substances quickly and accurately.

Wood basic density is an important indicator for assessing the pulping properties of raw wood materials. Rapidly determining the basic density of wood chips using near-infrared spectroscopy (NIRS) can provide basic theoretical data for developing and optimising pulp production processes. However, the source complicacy of raw material leads to high variability within the moisture content of wood chips. These fluctuations in the raw material make it difficult for the NIRS model to give a stable prediction performance. In this paper, the moisture desorption process of poplar chips was dynamically monitored by near-infrared spectroscopy. Principal component analysis (PCA) was applied to distinguish the spectral features due to moisture content to explore the change of free water and bound water in wood fiber. In order to investigate the effect of moisture content on the NIRS prediction of wood density, the partial least square calibration (PLS) models were built using wood chips with different moisture content conditions, respectively. And then external parameter orthogonalization algorithm (EPO) was used to improve the robustness of predictive models by eliminating the influence of chip moisture. The results showed that the best prediction accuracy was obtained from water-saturated chips spectra due to full access to information about fiber structures. However, much water absorption information in the spectra was redundant and useless for modeling, and the variations in moisture content also led to unstable prediction performance. The spectral moisture correction based on EPO was an effective method for desensitizing the calibration model to the influence of moisture content, enabling robust and accurate prediction of basic density. The EPO-PLS model provided a performance with a root mean square error (RMSE) of 12.23 kg·m-3, determination coefficients (R2) of 0.883 4, and residual prediction deviation (RPD) of 2.93 under different moisture content. This study built a robust NIR calibration model which was robustified against the influence of the variations in moisture content on the wood density prediction. This technology may facilitate the expansion of potential applications of NIR spectroscopy in the paper and pulp industry.

With the emergence of high-speed analog-to-digital converters, shaping algorithms for handling digital pulses was applied broadly in nuclear instruments. Trapezoidal shaper and Gaussian shaper are the common shaping algorithms. The former can increase the pulse passing rate due to its narrowed output pulse. The latter can make better energy resolution of nuclear instruments because of its attractive signal-to-noise ratio (SNR). The output gaussian shaper pulse has to narrow to add the discernment ability to piled-up pulses. The recurrence formula of the symmetrical gaussian shaping algorithm (SGSA) is deduced based on Fourier transform in the assumption that the shape of the output pulse obeys the gaussian function and its amplitude is equal to the corresponding input pulse. Moreover, its recurrence formula is composed of only multiplier and adder, which are easily implemented in field programmable gate arrays. After dealing with ideal pulses with SGSA, the result shows that the discernment ability to piled-up pulses is weaker along with the bigger σ (standard deviation) of its output pulse. Compare the shape of the output pulse of the trapezoidal shaper at the smallest rise and flat-top time with this method under σ=3Ts. This proves this method has a better discernment ability to piled-up pulses. After dealing 180 000 simulating pulses under different SNR with SGSA, the result reflects that the average extractive amplitude increases at SNR<55 dB, and then maintain stability. Meanwhile, the output pulse is smoother after the bigger value of σ. Their frequency spectrums also prove it. The cutoff-frequency decreases, and the denoising effect rises when σ increases. It can increase the energy resolution of nuclear instruments. Applying the total-reflection X-ray fluorescence spectrometer shows that the energy resolution increases 5.42 % and total count rate decreases 13.02% when σ increases. The experiments further strengthen the above-simulating conclusion. Those results proved that SGSA improves the discernment ability to piled-up pulses and can be implemented high-speed real-time processing for a pulse on a multi-channel analyzer.

The tensile strength is an important index to assess the mechanical properties of the wood. In order to solve the problems of low model accuracy caused by the small samples and redundant wavelength information in near-infrared spectroscopy modeling, a novel method combining wavelength optimization of MC-UVE-IVSO and PLS is proposed to predict the wood tensile strength. Firstly, 150 birch samples were selected as experimental objects, and the near-infrared spectrometer in the band of 900~1 700 nm was used to collect the spectral data of the test specimens, and the true tensile strength values were obtained by the mechanical testing machine. Secondly, the collected spectral data were preprocessed to complete smoothing filtering by combining multivariate scattering correction (MSC), first-order derivation and convolution smoothing (SG). Thirdly, the optimization methods, which include the variable combination cluster analysis algorithm (VCPA), the Monte Carlo uninformative variable elimination method (MC-UVE), the iterative variable subset optimization algorithm (IVSO) and the MC-UVE-IVSO combined optimization algorithm, were applied to select spectral wavelength features, and the optimal wavelength results based on different method were compared. Finally, the partial least squares birch tensile strength prediction model was established based on the selected wavelength of MC-UVE-IVSO. The experimental results show that the number of spectral variables is reduced from 512 to 98 based on the MC-UVE-IVSO and PLS, and the selected wavelength features account for 19% of the total wavelength. The predicted coefficient of determination (R2) is 0.940 4. The root mean square error of prediction (RMSEP) is 12.370 7. The ratio of performance to deviation (RPD) is 3.162 4, compared with full band, MC-UVE, VCPA,MC-UVE-VCPA and IVSO, R2 indicators (0.926 5, 0.828 2, 0.931 7, 0.934 3), RMSEP indicators (13.910 5, 17.355 2, 13.402 8, 14.070 5) and RPD indicators (2.812 3, 2.254 1, 2.918 8, 2.780 3) have been improved to varying degrees; In addition, the box plot of the prediction model established by statistical characteristic wavelengths further proves the stability of the MC-UVE-IVSO algorithm in dealing with multivariate wavelengths. The experimental results proved that the MC-UVE method could eliminate most of the variables, which are not related to the model, and the IVSO algorithm can effectively search for the optimal subset of variables. Based on the MC-UVE-IVSO optimization algorithm, the combination method has complementary advantages, and the optimized features can improve the accuracy and stability of the birch tensile strength prediction model. The method provides a theoretical basis for Non-destructive testing of wood samples based on near-infrared spectroscopy.

To find the differential chemical indicators between the raw and milk-processed corals, which can provide references for improving the quality standards of the milk-processed corals. A serial of the milk-processed corals was prepared by the orthogonal design of the processing factors, including the milk-to-water ratio, the decocting temperature, the drying time and the drying temperature. Calcium carbonates content and crystal form in the raw and milk-processed corals were tested by titrimetric analysis and X-ray diffraction, respectively. The organic and inorganic compositions of the raw and milk-processed corals were characterized by infrared spectroscopy.The mass concentration of calcium carbonatewas about 80% in the raw and milk-processed corals, while the calcium carbonate content could increase slightly when the raw coral was processed according to the processing specification. The crystal form of calcium carbonatewas calcite in the raw and milk-processed corals, while the X-ray diffraction peak of the raw coral at 2θ=47.953° could change after the processing. The organic compositions were quite different between the raw and milk-processed corals. The characteristic bands in the regions of 3 050~2 750, 1 770~1 720 and 1 710~1 600 cm-1 (corresponding to fats, proteins, carbohydrates, etc.) in the infrared spectra of the milk-processed corals were significantly stronger than those in the infrared spectrum of the raw coral. The areas of these infrared spectral bands were also sensitive to the variations of the processing factors, including the milk-to-water ratio, the decocting temperature, the drying time and the drying temperature. Infrared spectroscopy can objectively and quantitatively discriminate the raw and milk-processed corals. It was also possible to use infrared spectroscopy to evaluate the compliance of the processing methods of the milk-processed corals.

The heavy metal lead (Pb) pollution in water impacts human health and the water ecological environment. In order to realize the on-site and rapid detection of heavy metal Pb in water, in this paper, Chlorella pyrenoidosa was used as the adsorbent, and the rapid detection of heavy metal Pb in water based on enrichment by Chlorella pyrenoidosa combined with X-ray fluorescence (XRF) spectroscopy was carried out. The results show that when the pH value of the reaction solution of Chlorella pyrenoidosa and heavy metal Pb was 7. The reaction temperature was 25 ℃, Chlorella pyrenoidosa had the fast and high-efficient adsorption characteristics to heavy metal Pb, when the reaction time was 5 min, the adsorption efficiency of heavymetal Pb in the wide concentration range of 0.012 8～0.353 5 mg·L-1 was as high as 92%, but the adsorption efficiency of metalloid As was lower than 5%. Therefore, the enrichment based on Chlorella pyrenoidosa could effectively avoid the interference and influence of the optimal Kα characteristic peak of As on the optimal Lα characteristic peak of Pb in the XRF measurement process when the heavy metal Pb and metalloid As coexist; Under the optimal adsorption reaction conditions of Chlorella pyrenoidosa for heavy metal Pb, when the enrichment volume of the reaction solution was 10 mL, a quantitative detection method of heavy metal Pb in water based on the combination of Chlorella pyrenoidosa enrichment and XRF spectroscopy was established. There was a good linear relationship between the concentration of heavy metal Pb in water and the net integrated fluorescence intensity of the Lα characteristic peak of Pb in the XRF spectrum with a correlation coefficient r of 0.990. The detection limit of the method was 7.2 μg·L-1, which was lower than the standard limit of heavy metal Pb in the Class Ⅰ water quality standard in “Environmental Quality Standard for Surface Water (GB 3838—2002)” of China. This method was adopted to detect the heavy metal Pb in the actual water samples of Paihe, Kuanghe, Nanfeihe, Silihe and Shiwulihe in Hefei City, and the recoveries were all within the range of 87.84% to 115.66%, indicating that the established rapid detection method of heavy metal Pb in water which combined with enrichment by algal cells and XRF spectroscopy could be well applied to the rapid analysis and detection of heavy metal Pb in actual water. This research will provide a method basis for developing on-site and rapid monitoring techniques and instruments for heavy metals in water.

The aerodynamic levitator laser (ADL) heating device coupled with picosecond time-gate Raman spectrometer was built. It breaks through the limitation of temperature and crucible material of conventional heating method. It greatly shields the interference of blackbody radiation on Raman signal under high temperature and extreme conditions by relying on the extremely short measurement cycle of picosecond pulsed laser. In-situ Raman spectra of MgTi2O5 melt with high melting points at superhigh temperatures (1 903, 1 953, 2 003 K) were measured for the first time. In-situ temperature-dependent Raman spectra of MgTi2O5 crystal before melting (1 673 K) were measured by coupling the third-generation intensified charge-coupled device (ICCD) detector and nanosecond pulsed laser. Raman spectra of the crystal broaden and redshift with increasing temperature from room temperature (RT) to 1 953 K, and the relative intensity decreases. A single broad envelop was observed when the temperature was increased to the melt (2 003 K). Indicating that the long-range ordered structure of the crystal has been destroyed and the microstructures in the system have changed essentially. The Raman spectrum of MgTi2O5 crystal at RT was calculated by density functional theory (DFT), and major vibration modes were thus assigned by comparing the calculated spectrum with the experimental one. The vibration peaks in the low wavenumber region (<350 cm-1) can be mainly attributed to the external lattice vibration modes. The peak at 485 cm-1 in the medium wavenumber region corresponds to Ti—O—Ti bending vibration, and the main characteristic peak at 648 and 787 cm-1 stand for O—Ti stretching vibration and O—Ti—O bending vibration in the TiO6 octahedron, respectively. A series of cluster models assumed in melt were constructed and simulated by the quantum chemistry ab initio calculation method. The characteristic Raman active vibration wavenumbers and their scattering cross section were obtained. After the experimental Raman spectra of melt were corrected by scattering cross section, the deconvolution of the molten Raman spectra was carried out, and the concentration distribution of various species was thus described quantitatively. Results show that there are TiO4 tetrahedral clusters (The relative mole fractions of respective species Qi in different configurations are 54.6%Q0, 20.1%Q1, 5.0%Q2 and 4.8%Q3, and Qi is the titanium oxide tetrahedron with different bridge oxygen number i) and TiO6 octahedral clusters (Hexacoordinated titanium oxide octahedron, whose relative mole fraction is 14.8% H0) in MgTi2O5 melt. Ti4+ mainly exists in isolated tetrahedral structure Q0 and dimer structure Q1, and a small part exists in the form of isolated titanium oxide octahedral H0. The isolated structure accounts for most of the composition of MgTi2O5 melt, which destroys the systems network connectivity and inhibits the glass forming ability. No solid-solid phase transitions were observed for MgTi2O5 crystal with the increasing temperature before melting. Above the melting point, the Ti—O polyhedron in the crystal changes from a single TiO6 species to the coexistence of both TiO4 and TiO6 species.

Xitou Site, located in Xunyi County, Xianyang City, Shaanxi Province, contains rich human cultural relics from the Neolithic Age to the Ming and Qing Dynasties. In the excavation sites of Shangxitou Village and Nantou Village, many cave-dwelling sites of the Neolithic Age were found, and the floor of some of these sites was a Composite building material of “Baihuimian” (Lime layer) and organic-tempered daub, which were relatively well preserved. This paper used stereoscopic microscope, polarizing microscope, ultra-depth of field microscope and scanning electron microscope(SEM) to observe the structure of “Baihuimian and organic-tempered daub” floor. In order to further solve the problems of sample production materials, production process analysis and technical principle, the means of X-ray Diffractometer(XRD), Fourier transform infrared spectrometer (FT-IR), thermal analysis (TGA-DSC), energy dispersive spectrometer (EDS) were used to analyze and detect “Baihuimian”, organic-tempered daub and “Baihuimian and organic-tempered daub” floor. The results showed that the main ingredients of “Baihuimian” are silica dioxide and calcium carbonate obtained by carbonization of loess-doll after calcination; The “Baihuimian” has three layers, which should be smeared three times, and the thickness of each layer of “Baihuimian” is the same and not more than 1 mm; The substrate of “Baihuimian” is organic-tempered daub, in which the straw fiber can strengthen and inhibit cracking. When daubing the “Baihuimian”, the lime water infiltrated into the organic-tempered daub and calcium hydroxide itself was carbonated to produce calcium carbonate, contributing to organic-tempered daub structures stability and strength. The two building materials are still closely combined, with only a 10~20 μm distance between them, which indicates that at least in the late Neolithic Age, Chinese ancestors had widely mastered the compound building material and its construction technology. This study is of great significance to understanding and utilising natural materials by ancient Chinese ancestors and to sort out the development of traditional architectural science and technology civilization.

The classification of stars has always been a hot topic in recent astronomical research. The classification of stellar subtypes is significant for exploring stellar evolution and identifying rare celestial bodies. This paper designs the SSTransformer (Stellar Spectrum Transformer) classification model for the LAMOST spectral subtype classification problem. The model mainly comprises three parts, including the input module, the embedding module, and the SST encoding module. In the input module, the spectral data is processed into blocks, which are mapped to vectors through a linear projection layer. In the embedding module, in order to extract useful data features, the output of the linear projection layer is added to a learnable category embedding block. In order to preserve the position information, a position embedding block is added, and then these data feature vectors are sent to the SST encoding module. Finally, the data features are extracted in the SST coding module, and the stellar spectrum is classified using the multilayer perceptron combined with the new features. In this paper,the spectral data of type A, F, G, K, and M starsis all from the one-dimensional low-resolution spectra in LAMOST DR8, 35256 pieces of one-dimensional spectral data are used for training the SSTransformer model, and 8 815 pieces of one-dimensional spectral data are used for testing the SSTransformer model. In order to speed up the convergence of the model, Z-Score normalization is used for the data. Because this is a multi-classification problem, the experiment adopts five evaluation indicators: accuracy rate, precision rate, recall rate, F1-Score, and Kappa coefficient. The experimental results show that the SSTransformer model can effectively screen and classify one-dimensional stellar spectral data, and the classification accuracy reaches 98.36%, which is higher than the support vector machine (SVM) algorithm, eXtreme Gradient Boosting (XGBoost) algorithm, and convolutional neural networks (CNN).

Redshift is one of the basic parameters of galaxies. The number of photometric images is enormous relative to the spectrum. A large number of known galaxies have only photometric images and no spectra. It is, therefore, common to obtain redshift values from photometric images rather than spectra. This paper constructs the Galactic Redshift Regression Network (GRRnet), a convolutional neural network for estimating photometric redshift from galaxy images. It has a deeper network layer than previous similar methods and adds an attention mechanism to help it focus on more useful information. Based on GRRnet, this paper further proposes a two-step strategy, GRRnet-C-R: the first step is to classify the galaxies according to the redshift roughly; the second step is to perform regression estimation according to the classified categories, and finally merge them. This strategy can significantly reduce the error of photometric redshift estimation. The data in this paper are all from the Sloan Digital Sky SurveyDR16. The relevant data of each galaxy includes the composite images of the three bands of g, r, and z, the photometric values of the five bands of u, g, r, i, and z, and the observed Spectral redshift for labeling. In the pre-processing process, the photometric image is cut to the size of 50×50, to ensure that most of the galaxies can be framed while the computation is reduced. Since the input size of the comparison algorithm NetZ is 64×64, to keep the input size consistent, use the function cv2. resize to change the image size to 64×64. The experimental results show that the mean squared error (MSE) of GRRnet-C-R reaches 0.001 46, which is 22.3%, 21.9% and 18.0% lower than that of random forest (RF), eXtreme Gradient Boosting (XGBoost) and NetZ, respectively. The linear regression coefficient of determination of GRRnet-C-R reached 0.948, which achieved a good model fitting effect. Experimental results show that this two-part strategy can effectively reduce the error of metering redshift estimation, which provides a new idea and method for subsequent metering redshift estimation.

Spectral solar radiometers can directly measure the variation of solar radiation and reflect the radiation information corresponding to each band. The calibration accuracy of the instrument in the whole band of the direct ray channel directly affects the inversion accuracy of atmospheric parameters. The commonly used Langley fitting method has low calibration accuracy in the atmospheric strong absorption band, and there are certain errors in the final calculation of precipitable water and strong absorption band transmittance data. In order to meet the requirements of high-precision measurement of the visible-near-infrared full-band solar spectrum, this paper proposes a Mixing calibration method combining the Langley calibration method of the non-absorption band and the Mixing calibration method based on theoretical calculation of solar irradiance at the top of the atmosphere of the strong absorption band. It obtains the calibration value of the full band of the spectral radiometer. Because the instrumental response function changes slowly with wavelength, the instrumental response function in the strong absorption band is obtained by linear interpolation of the calibration instrument response function in the non-absorption band according to wavelength. Then the instrumental response function in the strong absorption band is obtained by combining the relationship between the solar irradiance at the top of the atmosphere and the instrumental calibration value. By comparing the change curves of instrument calibration values of the Langley calibration method, improved Langley calibration method and Mixing calibration method, it is found that the calibration values of the former two methods have obvious mutations in the strong absorption band, while the calibration values of the Mixing calibration method change more gradually in the strong absorption band, which conforms to the law of instrument response. By comparing the relative deviation between the atmospheric transmittance measured by different calibration methods and the transmittance calculated by CART theory, it is found that the average deviation of the Mixing calibration method is reduced by 1.15%, which is mainly attributed to the improvement of the measurement accuracy of atmospheric transmittance in the strong absorption band by the Mixing calibration method. The precipitation data calculated by the improved Langley method and the Mixing calibration method are compared with those measured by the same type of POM radiometer abroad. The results calculated by the Mixing calibration method are almost consistent with those calculated by the POM radiometer, and the relative error is less than 10%. Compared with the improved Langley scaling method, it reduced by 40% on average. For the atmospheric transmittance measured, compared with the POM radiometer transmittance data, the relative error of the Mixing calibration measurement is reduced by 25% at the 940 nm vapor absorption zone. Therefore, the Mixing calibration method is of great application value to the full-band calibration of a direct channel of spectral solar radiometer, the calculation of precipitable water and the calculation of transmittance of the strong absorption band, and improves the calibration accuracy of the strong absorption band.

In recent years, a new kind of quartzose jade named “Mianlv Yu” from Myanmar has emerged in the west Yunnan jewelry market. It is characterized by delicate texture, and green color with different blue and yellow tones, and some green are similar to high-quality Australian green chalcedony. However, the color origin of this jade is still unclear, and there is still a lack of relevant theoretical support for its species identification, quality evaluation, and market promotion. In this paper, the mineral composition and structure, chemical constituent, spectral characteristics, and color origin of this jade were studied using an Infrared spectrometer, Raman spectrometer, UV-vis spectrometer, X-ray fluorescence spectrometer, X-ray powder diffractometer, and Polarizing microscope. The results show that the main mineral is α-quartz (containing trace Moganite), mainly cryptocrystalline, and a small amount of microcrystalline, accounting for more than 90% of the content, followed by sericite and willemsite in the form of microgranular and scaly, and extremely trace amounts of nepouite and chromceladonite. Occasionally, there is secondary disseminated iron clay in the local position, and the whole structure is microgranular with scales. The infrared transmission spectrum mainly shows the infrared absorption characteristics of α-quartz, Infrared absorption peaks at 1 019, 800～600, and 462 cm-1 are attributed to the anti-symmetric stretching vibration of νas(Si—O), and the symmetric stretching vibration of νs(Si—O—Si) and the bending vibration of δ(Si—O), respectively. The characteristic absorption peaks at 3 463, 1 639, and 1 399 cm-1 are caused by the anti-symmetric stretching vibration νas(H—O—H) and bending vibration δ(H—O—H) of free water molecules between quartz microvoids. In the Raman spectrum test, except for the Raman group peaks 204, 262, 355, 395, and 463 cm-1 indicated the characteristics of α-quartz, the weak Raman peak at 501 cm-1 indicated the presence of trace Moganite, and the Raman peak at 675 cm-1 indicated the presence of willemsite. The chemical composition and ultraviolet-visible spectra characteristics show that the jade contains trace impurity elements such as Mg, Al, Cl, K, Ca, Ti, Cr, Fe, Ni, Ni, and Fe are the main chromogenic elements. This jades significant difference in Ni and Fe content reveals why it presents two different color series: green to bluish green and greenish yellow to yellowish green. The high content of Ni and low content of Fe produces green to bluish green series, and the change of blue tone is positively correlated with the content of Ni. The same degree of low content of Ni and Fe produces greenish yellow to yellowish green series, and the change of yellow tone is negatively correlated with the content of Fe and Ni. In conclusion, the mineral species of “Mianlv Yu” is green chalcedony, and its color characteristics should be attributed to impurity minerals such as willemsite, sericite, and secondary iron argillaceous. Nickel exists in the forms of free Ni ions and impurity mineral willemsite. Willemsite is a rare mineral in other sources of green chalcedony, which should be considered an important reference feature for origin traceability. This study has enriched the data of green chalcedonys varieties and its origin information, and also provided essential data for further exploring the metallogenic geological background of “Mianlv Yu” Jade.

The Xuebaoding deposit which locates in Huya Township, Pingwu County, Mianyang City, Sichuan province, produces a kind of big rare orange scheelite with perfect crystal shape that collectors of gemstones and mineral crystals favor. The conventional gemological characteristics, infrared spectroscopy, Raman spectroscopy, ultraviolet-visible spectroscopy, and fluorescence spectroscopy were employed to study the spectral characteristics of five scheelite samples from Xuebaoding in this article in order to clarify the gemological characteristics of scheelite in this area . The typical infrared spectra show that: the fingerprint characteristics absorption in 440 cm-1 and region 800~900 cm-1 is induced by out-of-plane bending vibration andasymmetric stretching vibrations attributed to ［WO4］2- tetrahedral groups, respectively. The functional group region (2 000~3 000 cm-1) absorption peaks are related to water. The Raman spectrum scattering main peak at 911 cm-1 corresponds to the ν1 symmetric stretching vibration of ［WO4］2-; the Raman shift located in 797 cm-1 is caused by ν3 anti-symmetric stretching vibrations of ［WO4］2-; the low intensity Raman scattering peak at 322 and 400 cm-1 correspond to the out-of-plane flexural ν2 vibration of ［WO4］2-; The translational mode of (Ca—O) displays the Raman spectrum scattering peak at 211 cm-1. The ultraviolet-visible spectrum shows that the deep orange color of scheelite samples from the Xuebaoding deposit is related to absorption peaks at 584, 588, 682, 743, 750, 803 and 874 nm. The ultraviolet-visible spectrum may correspond to the “didymium” which is a compound of Pr and Nd. The 3D fluorescence spectra show that colorless scheelite sample and deep orange scheelite samples have the same main fluorescence peak position and number, which is located in λex235 nm/λem455 nm, λex250 nm/λem490 nm andλex265 nm/λem523 nm. In addition to the above-mentioned main fluorescence peaks, the light orange scheelite samples also appear to have a fluorescence peak at λex250 nm/λem425 nm.

Diaspore is a popular color gemstone in the jewelry market. It is popular with consumers because of its unique Alexandrite effect (brownish green in the sunlight and purplish red in the incandescent lamp). The study on the Alexandrite effect origin of diaspore is of great significance to the cutting, the treatment and the value evaluation of gemstones. There are few studies on the Alexandrite effect origin of diaspore. Considering that the chemical composition and crystal structure of diaspore are similar to corundum, the theory on color origin of corundum is relatively mature. Therefore, to research the Alexandrite effect of diaspore, the corundum, which is very similar to the Alexandrite effect of diaspore, is selected for innovative comparative study from trace elements, UV-Vis spectrum and crystal structure. The trace elements in the samples were measured by laser ablation inductively coupled plasma mass spectrometer. According to the test results, the main chromogenic elements in diaspore are Fe, Cr, V, Ti and the main chromogenic elements in corundum are Mg, Ti, Fe, V, Cr. The types of trace elements in the two samples are similar, but the contents are different. Ultraviolet-visible and polarized ultraviolet-visible spectra were used to characterize the absorption characteristics of the samples in the visible light region. It was found that there were absorption peaks at 387, 398, 438 and 448 nm and absorption bands with centers at about 572 nm in the spectrum of diaspore. It was this absorption band that caused the Alexandrite effect. Correspondingly, there were absorption peaks at 377, 388 and 450 nm and an absorption band with the center at about 560 nm in the spectrum of diaspore. Their absorption characteristics in the visible light region are very similar. The difference is that the absorption band at 560 nm of corundum does not have obvious polarization, while the absorption band at 572 nm of diaspore has polarization. The absorption characteristics caused by charge transfer in crystals often have polarization. Through the comparative analysis of their crystal structures and the charge compensation theory in corundum, we speculate that the absorption peak at 398 nm in diaspore is caused by the d—d electron transition of Fe3+, the absorption peaks at 387, 438, 448 nm are caused by the Fe3+-Fe3+ pair, and the absorption band at 572 nm is caused by the Cr, V, Fe2+-Ti4+ pair. The alexandrite effect in diaspore is caused by Cr, V, Fe2+-Ti4+ pair. This study innovatively studied the Alexandrite effect origin of diaspore by comparing it with corundum, providing a new idea for studying similar problems in gemstones.

Periodontitis is an infectious, destructive and inflammatory disease of periodontal tissue around teeth. The main clinical manifestations of periodontitis are soft tissue pocket formation, clinical attachment loss and alveolar bone resorption. Visible near-infrared spectroscopy, characterized by noninvasive and rapid detection, has been widely used in medicine. The present study explored the use of visible near-infrared spectroscopy in evaluating the relative contents of oxygenated hemoglobin and deoxyhemoglobin in gingival tissue of severe periodontitis. The gingival tissue spectra (400～1 700 nm) were obtained and processed from 20 sites of 5 healthy subjects and 20 sites of 5 patients with severe periodontitis. Spectra were collected at the gingival margin, 4 and 7 mm below the gingival margin. Our research found that oxygenated hemoglobin and deoxyhemoglobin showed obvious spectral absorption characteristics at 544 and 576 nm respectively. The relative absorption depths of oxygenated hemoglobin and deoxyhemoglobin were calculated from the continuum removal based on the original spectral data. The results showed that the relative contents of oxygenated hemoglobin and deoxyhemoglobin in the periodontal pocket of severe periodontitis were significantly higher than those in the healthy group (p<0.05). At the same time, there was no significant difference in the contents of oxygenated hemoglobin and deoxyhemoglobin at different depths in the deep periodontal pocket of severe periodontitis. The results reflected the hemodynamic differences between severe periodontitis and healthy gingival tissue, and provided a scientific basis for applying visible near-infrared spectroscopy in noninvasive detection and auxiliary diagnosis of periodontitis.

The soil organic content is the main basis for developing soil fertilization programs in jujube orchards. A reasonable fertilization program is of great significance for improving the quality of jujube, reducing farmers investment and increasing the output of jujube orchards . However, it is time-consuming and resource-intensive to obtain SOM content of jujube orchards using the traditional method, which does not meet the needs of precise fertilization management in jujube orchards. At the same time, the hyperspectral detection of soil organic matter is an effective alternative method. 158 soil samples are collected by grid distribution method, and the indoor hyperspectral data and SOM content of air-dried soil samples are determined. The 400~2 400 nm full waveband (R) and the datasets selected by three data reduction algorithms of competitive adaptive weighting algorithm (CARS), successive projection algorithm (SPA) and particle swarm optimization algorithm (PSO) are combined with three modeling methods, which are partial least squares regression (PLSR), back propagation neural network (BPNN) and convolutional neural network (CNN) to construct 12 combined inversion models of SOM content of jujube orchards. Moreover, the optimal spectral inversion model of SOM content of jujube orchards was selected by comparing the accuracy evaluation index and training time of the models. The results show that (1) CARS, SPA and PSO can all compress the spectral data to less than 10% of the original data, and the number of screened wavelengths is reduced from the original 2001 variables to 98, 156 and 102, respectively. The validation set RPD of the dimensionality reduction combined model are all greater than 1.50, and all of them can achieve the inversion of the SOM content of jujube orchards. Compared with the R combined model, the dimensionality reduction combined model can save at least 30% of time cost, especially the combined model constructed with BPNN and CNN can save 90% of the training time, and the model has stronger stability and better model effect. (2)The validation set of the CARS dataset to construct the combined model has R2 greater than 0.85 and RPD greater than 2.50, which is the best among the three-dimensionality reduction algorithms; the validation effect of the combined model of the PSO dataset is slightly lower than that of the CARS dataset, but better than that of the R dataset, with R2 greater than 0.80 and RPD greater than 2.00; the validation effect of the SPA dataset to construct the combined model is lower than that of the R dataset The validation effect of the SPA dataset is lower than that of the R dataset, and the effect is the worst among the three-dimensionality reduction algorithms. (3) Both BPNN and CNN methods outperformed the PLSR model in terms of inversion model validation, while the BPNN model out performed the CNN model in terms of model training time and model validation effect, and its validation effect combined with the CARS dataset is optimal with R2 of 0.91, PRD of 3.34, nRMSE% of 11.93, and training time of 58.00 s. The model can detect the SOM content of the jujube orchards rapidly. The CARS-BPNN model is the optimal model for the inversion of SOM in jujube orchards in South Xinjiang, and the results of the study can provide a reference for the rapid detection of soil nutrients and formulation of fertilization plan in jujube orchards in South Xinjiang.

Grapefruit peel is thick. The peel and pulp belong to two different media, and the refractive index and absorption of light from the peel and pulp are different. Modeling of grapefruit soluble solids content without removing the interference of fruit peel, can lead to poor model accuracy. To address the problem of poor model accuracy due to the mismatch between spectral acquisition and target when building a fruit quality detection model. To address the problem of poor model accuracy due to the mismatch between spectral acquisition and target when establishing fruit quality detection models, this study takes ShangraoMajiya grapefruit as the experimental object, builds an adjustable experimental platform independently, acquires and analyzes the light energy attenuation pattern of whole grapefruit, searches for the relationship between grapefruit thickness and light transmission, and explores the effects of peel thickness and light transmission depth on grapefruit SSC detection accuracy. Firstly, the transmitted light source was placed directly above the equatorial circle of the grapefruit, and the spectral intensity received by different regions of the equatorial grapefruit circle was counted. The spectral intensity distribution was plotted, and the results showed that the further away from the light source emission point, the lower the spectral intensity. The light intensity received at the incident point from far and near positions accounted for 33.40%, 2.90%, 0.50%, 0.40%, 0.20%. The absorption of light by the grapefruit peel was more obvious, and the scattered light energy accounted for a smaller proportion. Secondly, the slicing method was used to record the remaining thickness and the corresponding spectral intensity, and to draw the curve of the changing pattern of spectral intensity.The smaller the remaining thickness, the greater the spectral intensity when the thickness was 32.90 mm, the spectral intensity changed dramatically, when the fruit thickness was higher than 32.92 mm, the photon intensity received by the fruit was generally lower when the fruit was lower than 32.92 mm, the spectral intensity undertook a jump increase. Then the flesh, whole fruit and peel spectra were collected, and the SSC prediction model was developed using partial least squares, and the best prediction was obtained for the peeled flesh. Finally, the spectra were collected when the thickness of grapefruit flesh, peel + flesh was 40, 30, 20 and 10 mm, and the SSC prediction models with different thicknesses were established. The correlation coefficients of the prediction sets were 0.91, 0.89, 0.87 and 0.86 when the thickness of flesh was 20, 40, 60 and 80 mm respectively. The SSC prediction model had the highest accuracy when the flesh was at a transmission depth of 20 mm. The spectral transmission depth of peel+pulp was 20, 40, 60 and 80 mm, and the prediction set correlation coefficients were 0.78, 0.86, 0.93 and 0.84, respectively, with the best prediction at a transmission depth of 60 mm for peel+pulp.The results show that the difference in tissue composition inside the fruit skin and pulp affects the results of SSC prediction, but changing the transmission distance of visible/NIR inside the fruit can also optimize the model accuracy. This study reveals the diffuse transmission characteristics of visible/NIR in fruit tissue, which can provide a realistic basis for developing online sorting devices for the quality of thick-skinned fruits.

With the development of laser technology, laser warning technology has become the focus of research in various countries to achieve defense against various incoming lasers, and corresponding evasion and counterattack measurements are taken according to the parameters detected by the laser warning system. At present, the main parameters of laser warning detection include incoming laser azimuth, pitch angle, laser wavelength and laser pulse width. However, the existing laser warning system can not achieve multi-parameter simultaneous detection, and the absolute direction of the incoming laser cannot be obtained due to the narrower spectral range, limited field of view, and relative azimuth angle of detection. In this paper, a new method that broadband, large field of view, multi-parameter laser warning is proposed to realize the high-precision comprehensive measurement of the wavelength, absolute azimuth, absolute pitch angle, and pulse width of incoming laser in the range of 450~1 700 nm, which is mainly composed of laser pulse width measurement, absolute angle and laser wavelength measurement, control and data processing. The pulse width measurement module comprises an optical lens, multiband narrowband filter, and high-speed photodetector to realize the photoelectric conversion of the incoming pulsed laser. The absolute angle and laser wavelength measurement module that is composed of a grating, large field of view broadband achromatic lens, multiband narrowband filter, broadband area array detector, and a three-dimensional electronic compass can obtain the incoming laser wavelength, relative azimuth and relative pitch angle by the position of the first and zero diffraction spots. Then, combined with the measured direction angle, pitch angle and roll angle of the three-dimensional electronic compass, the incoming lasers three-dimensional absolute azimuth and absolute pitch angle are obtained. The multiband narrowband filter is mainly based on the strobe filter of several commonly used military laser wavelengths, which effectively filters out the influence of the background light and reduces the systems false warning and leakage warning. After the theoretical derivation and analysis of the measurement method and parameters, we design a broadband multi-parameter laser warning detection system prototype, verifying the experimental feasibility. The experimental results show that the systems azimuth and pitch angles can reach 120° and 96°, respectively. The angle measurement accuracy is better than 1°, the central wavelength measurement accuracy is better than 10 nm, and the pulse width measurement accuracy is better than 5 ns. This technology lays the foundation for the high-precision multi-parameter comprehensive detection of incoming lasers in the sea, land, air and space fields and is expected to improve the survivability in complex battlefields.

Near-infrared spectroscopy (NIRS) technology has certain feasibility in identifying crop seed varieties, but if the storage time of the seeds to be tested is different, the accuracy of the identification model will be affected. In order to reduce the influence of storage time on the recognition model and improve the models prediction ability, NIRS technology and image processing technology are combined to extract spectral features related to physiological and biochemical indicators of varieties and apparent image features related to varieties. In order to extract the optimal spectral features, an improved backward interval partial least squares (IM_BiPLS) spectral interval selection algorithm is proposed. Aiming at the problem that it is difficult to determine the number of segments of BiPLS, the algorithm changes the number of segments within a certain range and takes the ratio of the correlation coefficient of the model established by the combination interval obtained by each segment number and the root mean square error of cross-validation as the evaluation index. When the index is maximum, the band combination corresponding to the segment number is the best. The competitive adaptive reweighting method (CARS) removes the uninformative and collinear variables in the selected band of IM_BiPLS and further optimises the spectral features. In order to extract the apparent image features related to varieties, firstly, the image segmentation algorithm based on maximum entropy and double region marking is used to remove the regions of interest and segment the single seed image. Then a single seeds morphological, texture and color features are extracted, and the statistical average features of all seeds in each image sample are calculated. Finally, CARS are used to optimize these features to complete image feature extraction. Taking 10 yellow maize varieties as the research object, NIRS data and corresponding images of 216 samples were collected. For spectral data, use IM_BiPLS algorithm selects the band combination with 736 variables from 1845 variables in the full spectrum and uses CARS to optimize 29 spectral variables further; For image data, 29 image features are extracted, and 11 image features are further optimized by CARS. Respectively using the spectral feature band extracted by IM_BiPLS, the preferred feature wavelength extracted by IM_BiPLS_CARS, the image feature(image), the image feature extracted by CARS(image_CARS), and the fusion between IM_BiPLS_CARS and image_CARS(compound) as the input and the corresponding category of the sample as the output to set up BP neural network models. The test results show that the performance of the compound BP model is the best, the training accuracy and verification accuracy are 100%, and the test accuracy is 97.7%. The experimental results demonstrate that the fusion of NIRS features with image features can effectively improve the accuracy of the recognition model and reduce the impact of storage time on the model. This provide a reference method for achieving the non-destructive, rapid and accurate recognition of corn seed varieties.

The cucumber disease images acquired in natural scenes have noise, such as light and soil, which seriously affects the accuracy of cucumber disease recognition. The existing detection models occupy a large memory, making it difficult to achieve real-time detection of cucumber diseases.The visible spectral images of three diseases of cucumber, namely downy mildew, powdery mildew and anthracnose, in the natural environment are used as the research object. In this paper, a cucumber disease identification model based on the visible spectrum and an improved YOLOv5 object detection network is proposed to explore the accurate real-time detection of cucumber diseases in the natural environment and to reduce the storage cost of the detection model. The lightweight network structure YOLOv5s is used as the baseline model. The SE attention mechanism is introduced to extract the feature dimensional information to reduce the influence of complex background on the detection results and improve the detection accuracy of the model. The depth separable convolution is introduced to replace the standard convolution in the baseline model to reduce the computational burden caused by the model parameters and improve the detection speed. The network receives visible spectral images of arbitrary pixels and adjusts them to 640×640 pixels as input, outputs the cucumber disease occurrence region and disease category, initializes the detection method and trains the detection network using pre-trained weights on the COCO dataset.The experimental results show that the improved YOLOv5s-SE-DW model achieves 78.0%, 80.9%, and 83.6% detection accuracy for cucumber downy mildew, powdery mildew, and anthracnose, respectively, with mAP as high as 80.9%. The storage space of the model is only 9.45 MB, and the number of floating point operations is only 11.8 G. Compared with the baseline model, the mAP is improved by 2.4%, 4.6 G reduces the number of floating point operations, and the storage space required for the model is reduced by 4.95 MB. The improved model improves disease detection accuracy while reducing the storage memory. Further comparison with the classical two-stage target detection network Faster-RCNN and single-stage target detection networks YOLOv3, YOLOv3-tiny, YOLOv3-SPP, and YOLOv4 shows that the proposed YOLOv5s-SE-DW model improves the mAP by 3.8% compared with the best-performing YOLOv4 model among the comparison models, and the detection time and storage space are significantly reduced. The detection time and storage space are substantially reduced. The comprehensive results show that the proposed YOLOv5s-SE-DW network has good accuracy and real-time performance for cucumber disease detection in natural scenarios, which can meet the demand for disease detection in actual cucumber growing environments and provide a reference for cucumber disease detection in practical application scenarios.

Peoples pursuit of fish quality is getting higher and higher, so it is more and more important to develop the detection of the fat content of important parameters of fish in aquaculture. Although many researchers have modified and improved traditional detection methods, they are still time-consuming. It is laborious and requires professional personnel training. There are some problems. The emerging spectral technology also has problems such as low image quality caused by only using the whole fish fillet as a prediction sample, lack of universality, uneven distribution of components of the whole fish fillet, and long sampling time. This study uses the MCR-ALS algorithm. After reconstruction of the data and image gain, the feasibility of predicting and visualizing an important parameter (fat) of salmon fillets using near-infrared hyperspectral imaging was assessed. First, the fresh salmon bought from the market is cut into pieces according to the back and abdomen. Each salmon is made into 20 samples, a total of 100 samples, of which 75 samples are used for the calibration set, and 25 samples are used for the prediction set. Then, the spectral data of salmon fish samples were collected by hyperspectral imaging system, the content of salmon fat was measured by Soxhlet extractor, and the physical and chemical value samples were established. Then the spectral data was reconstructed by MCR-ALS. It was found that the reconstructed spectrally valid information increases with the component recommendation score, and then the characteristic wavelengths are selected by a continuous projection algorithm (SPA), and a least squares support vector machine (LS-SVM) model is established to evaluate the two prediction effects (raw and reconstructed data). MCR-ALS-SPA-LS-SVM has the highest prediction accuracy, Rp=0.955 5, RMESP=1.650 5, RPD=3.389 9. Then, using MCR-ALS and the unprocessed model to perform visual image prediction on fish fillet fat, its effect It greatly reduces the input of noise, effectively restores the outline of the fish fillet, and makes the fat stripes of the fish clearer, and the image quality is better than the latter. Further analysis of the cluster image, through the principal component contribution of different components and the principal component contribution ratio of the same component, it is found that when the category is 20, the sample will interfere with the background cluster, but it is found that only 5 and 10 A single species can completely describe the outline of the entire sample, and it has a good cluster presentation effect for spectrally strong reactive substances. It is possible to simplify the model. Whether it is data or images, the satisfactory prediction results confirm the feasibility of NIR hyperspectral imaging for salmon fat quantification and visual image prediction, and the optimization of the algorithm greatly shortens the detection time, creating better real-time online detection conditions.

In recent years, deep learning-based models have achieved remarkable results in the hyperspectral image (HSI) classification. Aiming at the low classification accuracy of deep learning-based HSI classification methods under limited sample data, this paper proposes an HSI classification method that combines ConvLSTM and a multi-attention mechanism network. The method is divided into three branches: spectral branch, spatial-X branch and spatial-Y branch to extract spectral features, spatial-X features and spatial-Y features respectively, and fuse the features in three directions for hyperspectral image classification. Since convolutional long short-term memory (ConvLSTM) shows good performance in learning valuable features and modeling long-term dependencies in spectral data, 3 hidden layers are used in the spectral branch, and the convolution kernel size is 3×3, the channels are 150, 100 and 60, respectively, to extract spectral information. On the spatial-X and spatial-Y branches, Dense spatial-X blocks and Dense spatial-Y blocks based on DenseNet and 3D-CNN are used to extract spatial-X and spatial-Y features, respectively. In order to enhance feature extraction, the attention mechanism of its feature direction is also introduced in these three branches, respectively. The spectral attention blocks are designed for the information-rich spectral bands, and a spatial-X attention block and a spatial-Y attention block are designed for the information-rich pixels, respectively. Experiments were conducted on three publicly available hyperspectral datasets, namely Indian Pines (IP), Pavia University (UP) and Salinas Valley (SV) datasets, and compared with five other methods: the SVM with RBF kernel (SVM), Going Deeper with Contextual CNN (CDCNN), Fast Dense Spectral-Spatial Convolution (FDSSC), Spectral-Spatial Residual Network (SSRN), Double-Branch Dual-Attention Mechanism Network (DBDA). In the experiments, the size of training and validation samples is set to 3% of the total samples on the IP dataset, and 0.5% of the total samples on the UP and SV datasets. For our method and all deep learning-based methods, the batch size is set to 16, the optimizer is set to Adam, the learning rate is set to 0.000 5, and the learning rate is dynamically adjusted. Since SVM directly uses spectral information for classification, the pixel size of the input sample block is 1×1, and the pixels of other input sample blocks based on deep learning methods are all set to 9×9. The experimental results show that the method in this paper can fully use the spectral and spatial characteristics of HSI, and achieve better results in the evaluation criteria such as OA, AA, and KAPPA. Among them, the OA index of the method in this paper is improved by 0.12%～2.04% on average compared with the suboptimal algorithm.

As an important part of the soil, the soil organic matter (SOM) can reflect soil fertility and quality. Compared with the traditional SOM measurement method, UAV hyperspectral images can quickly and real-time obtain the SOM content at the field-scale, which is of great significance for precision fertilization and sustainable utilization in the black soil region of Northeast China. In order to explore the difference in estimating the accuracy of SOM under crop cover by linear and nonlinear models based on hyperspectral data, the soil samples at the jointing stage and silking stage and UAV hyperspectral images were collected from the experimental corn field in the black soil region of Northeast China as the study area. The correlation between soil spectral reflectance and SOM content under crop cover was analyzed, and the spectral indices were calculated according to their response band. With the fertilizer rates and spectral indices as independent variables, multiple stepwise linear regression models (SMLR), support vector machine (SVM), random forest (RF) and eXtreme gradient boosting (XGBoost) models were established by screening characteristic variables, and the accuracies of the models were verified and compared (select R2 and RMSE as evaluation indicators). The results showed that the response band of SOM content under crop cover was 450~640 nm. Long-term application of chemical fertilizers had a significant effect on SOM content, and introducing it into the model as a covariate significantly improved the estimation accuracy of SOM. The test accuracies of the four models were: XGBoost>RF>SMLR>SVM, and the estimation result of XGBoost at the jointing stage was the best (R2 and RMSE of modeling set were 0.516, 0.253, and those of the verification set were 0.590, 0.222, respectively). Therefore, UAV hyperspectral technology can rapidly estimate SOM content in maize fields at field-scale, and the XGBoost model is a preferable option for estimating SOM content under crop cover conditions.

For the inadequate generalization ability of the quantitative evaluation model of crop yield, the lag of forecasting time and the difficulty of establishing the optimum lead yield estimation time, this paper takes Sentinel-2 remote sensing data and the measured maize yield as the data source to research the establishment of county-scale maize yield estimation and optimum lead yield estimation time. Based on the time-series image data of maize growth-satges, through building the correlation between maize yield measured data and vegetation index, the time-series maize yield estimation model was established by MLRM (multivariable linear regression model), GPR (Gaussian process regression model) and LSTM (Long short-term memory artificial neural network model). The experimental results show that LSTM is superior to GPR and MLRM in terms of the accuracy, and reliability of the yield prediction model, the capture of the abnormal yield value, and the optimum lead yield estimation time in the time series yield estimation model established with NDVI, GNDVI and GN ( NDVI and GNDVI combination ) as parameters. At the same time, based on the LSTM estimation model, the NDVI time-series image data up to tasseling stage were used as parameters and the yield prediction results showed that the R2(determination coefficient) was 0.83, RMSE(root mean square error) was 0.26 t·ha-1, RPD(relative percent deviation) was 3.52; The GNDVI time-series image data up to tasseling stage were used as parameters, and the yield prediction results showed that the R2 was 0.79, RMSE was 0.30 t·ha-1, RPD was 2.87; The GN time-series image data up to tasseling stage were used as parameters, and the yield prediction results showed that the R2 was 0.83, RMSE was 0.27 t·ha-1, RPD was 3.05. Using the NDVI time-series image data as the LSTM model parameter has the optimal yield estimation, and the maize yield could be predicted 2 months in advance compared with the maize harvest stage. As a result, we developed a crop yield forecasting method in this study to predict crop yield for county-scale. It has practical significance for maize yield forecasting and provides a relevant reference for similar crop yield estimation research.

Hyperspectral technology can provide nearly continuous spectral curves of ground objects, which has excellent potential for retrievingthe soils components. This paper investigates components retrieval from contaminated soil by hyperspectral technology. By so doing, it analyzes thesoil cadmium (Cd) concentration measured in the laboratory and the corresponding hyperspectral curvature data obtained at the same period, following whichthe retrieval model for the soil Cd concentration from the hyperspectral data in light with the (Deep Forest 2021, DF21) model is developed. In this study, the original spectrum(OS) data and the data processed by the Principal Component Analysis (PCA) are used as the models input parameters. Subsequently, two models, i.e., the OS-DF21 model based on the original spectral data and the PCA-DF21 model based on the PCA processed data, are established. The relationships between the input parameters and soil Cd concentration are respectively obtained by the OS-DF21 model and PCA-DF21 model. Then the soil Cd concentrationis estimated from the testing samples accordingly. To evaluate the retrieval performance, three indices, namely the coefficient of determination (R2), Root Mean Square Error (RMSE), and Residual Predictive Deviation (RPD) applied in this study. It is found that the OS-DF21 model has the best performance for the retrieval of soil Cd concentration, whose R2, RMSE, and RPD are 0.873, 0.120, and 2.892, respectively. In contrast, the PCA-DF21 model has arelatively lower retrieval accuracy, with R2, RMSE, and RPD being 0.779, 0.159, and 2.190, though the PCA can reduce the dimensionality of the spectral data. In this regard, the DF21 shows good retrieval performance and can be an essential supplementary method for soil heavy metal surveys in the study area and similar environmental regions.

Near-infrared hyperspectral imaging technology (NIR-HSI) was used to collect spectral image texture information to realize the discrimination of different bruise grades of Lingwu jujubes. 200 long jujube samples with bruising gradesⅠ, Ⅱ, Ⅲ, Ⅳ and Ⅴ were obtained by the bruising device, and the calibration set and prediction set were divided according to the ratio of 3∶1. Hyperspectral images of jujubes with different bruising grades were collected by NIR-HSI, the region of interest (ROI) was extracted using ENVI software and the average spectral value was calculated. Orthogonal signal correction (OSC), baseline, multiplicative scatter correction (MSC), moving average (MA), savitzky-golay (S-G) and de-trending were used to preprocess the original spectra and a partial least squares-discriminate analysis (PLS-DA) model was established; Variable combination population analysis (VCPA), uninformative variable elimination (UVE), competitive adaptive reweighted sampling (CARS) interval variable iterative space shrinkage approach (iVISSA) and successive projections algorithm (SPA) was used to extract the characteristic wavelengths based on the spectral data obtained by the optimal pretreatment method, and then the PLS-DA model was built. The hyperspectral image was subjected to masking and principal component analysis (PCA). Then the gray-level co-occurrence matrix (GLCM) was used to extract the texture parameters of the image with the highest principal component contribution rate, includingparameters of the angular second moment (ASM), entropy (ENT), contrast (CON), and correlation (COR), the PLS-DA models of data fusionwereestablished.The results showed that in the PLS-DA model of the original spectrum, the accuracies of the calibration set and prediction set were 89% and 86%; The PLS-DA model of the original spectrum based on de-trending preprocessing was the best, the accuracies of the calibration set and prediction set were both 90%, which were 1% and 4% higher than the original spectrum model, respectively. The De-trending-SPA-PLS-DA model based on the characteristic wavelength obtained the best results. The accuracies of the calibration set and prediction set were both 90%, which remained the same results as the optimal preprocessing model; The De-trending-SPA-COR fusion model obtained the best performance with 92% accuracy on both the calibration set and prediction set, which were 2% and 2% higher than the optimal spectral data model, respectively. Therefore, NIR-HSI,in combination with texture information, could realize rapid and non-destructive discrimination of Lingwu jujubes with different bruise grades.

In order to realize the rapid and effective detection of exogenous sucrose content in finished black tea, Fengqing large-leaved species tea was used as a research sample, and a quantitative prediction model for exogenous sucrose content in finished black tea was constructed by using near-infrared spectroscopy. First, near-infrared spectral data were collected during the production of finished black tea samples with different exogenous sucrose contents (0, 250, 500 and 750 g). When processing the data, in order to improve the prediction accuracy of the model, four different preprocessing methods, standard normal transformation (SNV), multivariate scattering correction (MSC), smoothing (Smooth) and centering (Center), were selected to reduce noise and establish partial least squares regression ( PLSR) model, according to the effect of the model, the best SNV preprocessing method was selected, the correction set correlation coefficient (Rc) was 0.907, the prediction set correlation coefficient (Rp) was 0.826, and the relative percent deviation (RPD) was 1.75. In order to reduce the impact of redundant information in the spectrum on the model operation speed, the competitive adaptive reweighted sampling (CARS), shuffled frog leaping algorithm(SFLA), variable combination population analysis iteratively retaining informative variables (VCPA-IRIV) and variable iterative space shrinkage algorithm (VISSA) to extract the characteristic wavelengths sensitive to sucrose from the SNV preprocessed spectrum. After the full spectrum and the selected characteristic wavelengths were dimensionally reduced by principal component analysis (PCA), linear PLSR and nonlinear support vector regression (SVR) and random forest (RF) quantitative prediction models were established respectively. The results show that after SNV preprocessing, the performance of the nonlinear SVR and RF models is better than that of the linear PLSR model, among which VCPA-IRIV-SVR is the optimal model, its Rc value is 0.950, Rp value is 0.924, and RPD value is 2.51. The research shows that near-infrared spectroscopy is feasible for the quantitative prediction of sucrose content in black tea processing, which provides a theoretical support for the non-destructive testing of black tea safety and quality.