The high-order harmonic component of a 1 GHz modulated sinusoidal wave optical signal is injected into the directly modulated laser, and the interaction between the intensity of the light field and the carrier concentration forms the photocurrent. The photocurrent is then converted into a microwave frequency multiplication signal at the radio frequency (RF) port of this directly modulated laser. In the experiment, the feedback control the directly modulated laser wavelength and thus the automatic locking of the laser are realized by the powe rvariance of the frequency multiplication signal detected at the RF port. The microwave frequency multiplication signals at 10 GHz and 12 GHz are generated without photodetectors. Compared with those by the injection locking scheme without an automatic locking loop, the microwave frequency multiplication signals by automatic locking are more stable. The power fluctuation is stable at 2 dB within 10 min, and the phase noise deterioration at 10 kHz is as little as 2.2 dB. The produced largest frequency doubling number of microwave frequency multiplication signals depends on the maximum modulation rate of the directly modulated laser.

Two kinds of ultra-broadband single polarization single mode (SPSM) square-lattice photonic crystal fibers with liquid crystal infiltrating are designed based on the mode cut-off method. The effects of the additional holes, the pitches in the fiber core, and the diameters of the liquid crystal filled holes on the SPSM operation are investigated with the full vector finite element method. The designed band-tunable ultra-broadband SPSM single-core photonic crystal fiber is obtained. Its SPSM operation can be tuned from 0.75 μm to 2.59 μm, and the widest SPSM bandwidth is 970 nm under a confinement loss of less than 0.1 dB·km -1. Moreover, the designed dual-core photonic crystal fiber is achieved with SPSM operation within the wavelength range from 0.98 μm to 1.74 μm, which can be applied in the wavelength division demultiplexer. In addition, a photonic crystal fiber wavelength splitter with a length of only 1.06 mm is fabricated, which can be used to realize the splitting of 1.31 μm and 1.55 μm lights with SPSM operation.

In the conventional high-frequency millimeter wave system, the high-order sidebands can be directly ignored when the sideband suppression ratio is relatively large. However, for a small sideband suppression ratio, the big power of a high-order sideband obviously influences the generation of millimeter waves. The effects of high-order sidebands on the generated high-frequency millimeter waves are mainly demonstrated in this paper. In addition, an eightfold-frequency millimeter wave is obtained based on even-order sideband suppression by the cascaded Mach-Zehnder modulators (MZMs). In an actual optical millimeter wave system, the redundant low-order sidebands usually occur after the modulation of high-order sidebands, which makes the power of low-order sidebands accumulated. However the phases are incoherent, and thus the system is unable to demodulate. Therefore, the influences of high-order sidebands outputted from the modulators with different modulation depths are discussed in detail. From the analysis of high-order sidebands, one can know that when the radio-frequency signals satisfy the optimal phase formula, the ±1 st, ±3 rd, ±5 th, ±7 th order flat sidebands are generated by the first MZM, and the eightfold-frequency millimeter wave with a 4th order sideband can thus be obtained through the second MZM without a filter. With this scheme, a pure eightfold-frequency millimeter wave can be obtained without large radio-frequency signals. For this scheme, not only the design structure is simple, but also the spectral purity is high.

The hyperspectral super-resolution image set is obtained with the classical super-resolution method and the characteristics of these images are studied. A quality assessment method of hyperspectral super-resolution images is proposed based on three types of image feature vectors. In this method, the spatial natural statistics, the local frequency features and the local binary gradient of images are calculated, respectively, and three kinds of feature vectors are obtained. The regression forest model is established for the three types of low-level feature vectors to predict the image quality scores. Compared with other classical methods, the proposed algorithm possesses high accuracy and good subjective and objective consistency.

A high dynamic range image synthesis method based on camera array is proposed as for the characteristics of space target images in the sky background. The camera array system is used to obtain different exposure images in the same spatial area and the star centroid is used as the control point. Thus the registration of images among different cameras is realized. Combined with the inverse response function of each camera, the weight function is reasonably constructed, and the different exposure images after registration are thus combined into a high dynamic range image. The experimental results indicate that the dynamic range of the synthesized image is increased. The proposed method can be used to effectively improve the signal-to-noise ratio of images, which overcomes the problem of over-damping and over-saturation of target images and is conducive to the detection and extraction of targets.

The dense and low-resolution face is difficult to be detected under the influence of attitude, occlusion and scale change. We propose a context-sensitive multi-scale face detection (CSMS) method to solve this problem. First, the CSMS method introduces an extraction module which combines the face context information to enrich the discriminant information by effectively fusing the features of multiple receptive fields. Secondly, from the point of view of model structure design, the CSMS method uses multi-scale features to extract scale-specific feature vectors and achieve the robust scale variety in face detection. In the training phase, the CSMS method adopts the end-to-end learning method, and introduces the training method focusing on the hard negative examples to solve the class imbalance problem in the small-scale target detection, and improves the ability of the network to distinguish the difficult examples. Experimental results show that the proposed method is robust in unconstrained environments and achieves advanced detection performance on the Wider Face dataset.

To improve the qualities of underwater laser images, the generator network of the generative adversarial network is changed to be a deep convolution neural network which contains the jumping structure and the dilated convolution. The network is used to learn the end-to-end parameters which map the unrepaired images to the target images from the self-built data set and repair underwater laser images with strong backscatter light. Compared with the experiment results of the joint processing of the classic denoising method and the contrast enhancement method, the proposed method can fill and repair the backscattered light area. The peak signal to noise ratio obtained by the proposed method increases by an average of 9.10 dB and the feature similarity increases by an average of 0.11. The denoising, enhancing contrast, improving the non-uniform illumination are achieved and the backscattered light is removed well.

The terahertz time domain spectroscopy (THz-TDS) is used to quantitatively analyze the benzoic acid content in arrowroot, and the arrowroot is with a plastic bag. The original data are subjected to the pretreatments of multiple scattering correction, baseline correction, first derivative, second derivative, and so on. The partial least square (PLS) method is used to establish a prediction model for the detection of benzoic acid content in arrowroot. The detection results show that model-determining factor for samples without plastic bags is 0.975 and the predicted root mean square error (RMSEP) is 1.126%. In contrast, that for samples with plastic bags is 0.976 and the RMSEP is 1.356%.

A phase correction algorithm is proposed based on the structured light multifrequency phase-shifting technique. The phase points in the effective measurement area of coal are divided into two types and corrected in turn according to the reliability criterion constructed based on phase gradient. The applicability and measurement accuracy of the proposed algorithm are tested experimentally. The experimental results show that the proposed algorithm can be used to well resove the problem of invalid phase points occurred during the surface profile measurement of coal.

The workspace measurement and positioning system is a distributed measurement system based on the rotary laser scanning technology. The laser signals are always disturbed in the processes of generation from the transmitter, transmittance through the physical space and reception by the receiver, and simultaneously the measurement errors are introduced. The time signal error analysis method and the corresponding filtering technology are introduced, and the error components occurred in the measuring process are analyzed based on the laser signal timing principle. Meanwhile, the time series signals are preprocessed as well the time and frequency domain characteristics are analyzed based on the time series statistical theory. The common-mode errors are found existing in the time series data, and the principal component analysis method is adopted to eliminate them. The experimental results show that the noise amplitudes of different time series are reduced by 10% after principal component analysis filtering.

The laser ultrasonic visualization technology is used to study the reflection and transmission of pulsed laser excited surface waves on the defects with different depths. The visualized sound field of the surface wave at the defects is obtained experimentally. The research results show that as the defect depth increases, the reflection coefficient of surface wave first increases and then changes steadily, and however the transmission coefficient first decreases and then changes smoothly. These research results lay an experimental foundation for the quantitative detection of cracks by laser ultrasonic surface waves.

Two filling schemes based on new sub-aperture shapes are proposed, which can achieve a higher area filling ratio than that for the filling scheme based on a hexagonal close-packed array with traditional circular sub-apertures. By the evaluation of the far-field beam quality and the balance between the energy loss caused by beam truncation and the beam quality improvement by filling factor improvement, an optimal solution is obtained. The simulation results show that these two new schemes can increase the proportion of power in the far field bucket by 9.2% and 13.9% comparing with those based on a hexagonal close-packed array with circular sub-apertures. The proposed aperture filling optimization schemes have some guidance value for increasing the efficiency of tiled-aperture coherent beam combining of fiber laser arrays.

The feature difference among the images of fine-grained vehicle models is small and there exist many factors disturbing recognition under complex image background. To improve the feature extraction ability and the recognition accuracy of images under complex background, a classifier named Softmax-SVM is proposed based on deep convolutional neural network (DCNN) and support vector machine(SVM), in which the cross-entropy cost function is combined with the hinge loss function to replace the Softmax function layer, so that the over-fitting is avoided. Meanwhile, a 10-layer DCNN is designed to extract features automatically and the problem of manual extraction of features is also avoided. The experimental dataset consists of the images of 27 types of fine-gained vehicle models under complex background, especially of the similar models from the same car manufacturer. The experimental results show that the Softmax-SVM classifier can be used to recognize the 269 sample images without much emphasis on the pre-processing stages, and in the identification process, the accuracy rate is 97.78% and the time to identity each image is 0.759 s. The above model performs more efficiently than the traditional recognition methods and the unimproved DCNN models. In consequence, the Softmax-SVM classifier based on DCNN can adapt to the complex changes of environment and give consideration to both the recognition accuracy and efficiency, which provides practical reference value in the classification field of fine-gained vehicle models under complex background.

The Mask R-CNN (mask region-based convolutional neural network) object detection method is proposed based on the improved feature pyramid. The experimental results show that compared with the Mask R-CNN detection structure, the mean average precision (mAP) under different Intersection-over-Union (IoU) thresholds increases by 2.4% and 3.8% in the detection of object edge and bounding box, respectively. In particular, the detection accuracy of medium size objects is greatly improved by 7.7% and 8.5%, respectively, which indicates strong robustness.

A weighted iterative pose estimation method is proposed, which combines the object-space errors and image-space collinearity errors. Based on the efficient perspective-n-point (EPnP) algorithm, an improved iterative algorithm with adaptive weighted object-space collinearity error function is added to optimize the results, and the average image-space collinearity error after iteration is used as the minimum error decision threshold to correct the parameters with relatively large errors. The experimental results show that this proposed method can greatly improve the accuracy and robustness of the pose parameters of the EPnP algorithm in the non-redundant situations. Moreover, it can maintain the high precision of the EPnP algorithm and possess high practical value in the condition of sufficient reference points.

A scheme for designing polycarbonate (PC)/polymethyl methacrylate (PMMA) polymer alternating multilayer optical films with high reflectivity in near-infrared region (780~1100 nm) is proposed. The feasibility for the fabrication of the special multilayer optical films by the micro-nano laminated co-extrusion device is demonstrated. The optimal four-periodic Bragg center wavelength is retrieved by the genetic algorithm, and the optimum film thickness is obtained. The spectral characteristics of four-periodic superimposed multilayer films are simulated and the effects of thickness error and incident angle on the total reflectivity of the multilayer film in the near-infrared region are discussed. The results show that the proposed scheme can meet the requirements of solar heat insulation and intelligent interior lighting.

Based on the effective medium theory, the hyperbolic dispersion relationship of graphene-dielectric multilayer (GDM) hyperbolic metamaterials (HMMs) in the near-infrared waveband is analyzed. The transfer matrix method (TMM) is optimized. The transmission spectra of GDM-HMMs with different numbers of periods are calculated and analyzed. Based on the Fabry-Perot (F-P) cavity theory, the evolution of transmission spectra is theoretically analyzed, and the hyperbolic dispersion relationship of GDM-HMMs in the near-infrared waveband is verified. The research results show that the large tangential wave vector condition is needed for the realization of electromagnetic wave transmission in GDM-HMMs. The transmission spectral characteristics are influenced by the total number of structural periods, and the F-P cavity theory can be used for the analysis and reverse structural design.

Based on the tunable conductivity of semiconductor silicon, a controllable electromagnetic induced transparency (EIT) structure composed of cut wires (CW) and split-ring resonators (SRR) is designed, and the active modulation of EIT effect is realized. It is found that a narrow transparency window with a transmittance of about 94% appears in the transmission spectrum at near 1.33 THz when the conductivity of semiconductor silicon was 1 S/m. When the conductivity increases to 5000 S/m, the transmittance becomes 58% and when the conductivity approaches 15000 S/m, the EIT effect almost disappears, and the modulation efficiency approaches 66%. The transmission spectra under different conductivities are consistent with their fitting curves by the coupled mode theory, indicating that the simulation results are in accord with the theoretical calculation findings. Both the simulation and calculation results show that the damping ratio of dark mode and the loss increase when the electrical conductivity of silicon increases. When the electrical conductivity reaches a certain value, the resonance of dark mode is not be stimulated and thus the EIT effect disappears.

Taking the blood glucose concentration as an example, the accuracy of the blood glucose measurement system is improved by means of the simultaneous incorportation of the extraction data form dynamic spectra and the influencing factors of non-measured components into the prediction model. The blood glucose prediction model is established through the support vector machine algorithm. The modeling results show that the prediction value from the multi-factor-considered model is superior to that from the non-measurement-component-considered model. The correlation coefficient of the former is 0.9627, higher by 14.23% , the root mean square error is 0.13, reduced by 43.12%, and the number of samples with a relative error in the range of 10% is higher by 8.33%, compared with those of the latter.

The tunable optical narrow-band filter is one of the key components in an optical system, which has a wide application in many fields such as optical signal processing and optical sensing. A novel scheme for a widely tunable narrow-band filter is proposed using the micro-optofluidic technology and based on a sub-wavelength grating structure, in which a wide-range adjustable narrow-band filtering function is achieved by adjusting the refractive index of mixed liquid in the micro-fluidic channel. The optical performance is numerically investigated using the rigorous mode theory and the simulation results show that the tunable wavelength range of the designed device is over 28 nm, the filtering bandwidth is less than 0.03 nm, the sensitivity S is over 350 nm·RIU -1, and the quality factor Q is over 50000, indicating that the performance is superior. Moreover, the proposed novel device has many advantages such as simple structure, ease for fabrication and large fabrication tolerance, which is very useful in the fields of bio-sensing, chemical analysis and others.

A scheme for the preparation of quantum entangled states is proposed based on a V-type three-level atom and cavity quantum electrodynamics system. The dynamics of cavity-cavity entanglement is numerically analyzed under different coupling parameters. The research results show that the cavity-cavity entanglement can be mediated by varying the coupling parameters between cavities when the atom successively passes through the empty cavities at a constant speed. Moreover, this scheme can be utilized for the simultaneous generation and control of multi-component quantum entanglement states.

Based on the dynamic optical path delay device with different positions and the entangled source signals at Roland C main station as well as the phase difference information of entangled light fields with different paths detected by the main station unbalanced M-Z (Mach-Zehnder) interferometer, the synchronization time difference information of Roland C main and auxiliary stations is obtained. The theoretical and simulation results show that the optimal time difference information with a precision of tens of picoseconds is obtained when the phase difference is 0.5π. Compare with the original master-slave synchronization, free synchronization and others, the proposed scheme does not need to measure the pulse arrival time. Moreover it can break through the quantum noise limit and effectively improve the time measurement accuracy.

Anomaly detection plays a more and more important role in the hyperspectral image (HIS) processing field. Since the low-rank and sparse matrix decomposition (LRaSMD) algorithm can separate the anomalies from the background, it can protect the background model from corruption by anomalies and noises. A novel hyperspectral anomaly detection algorithm is proposed based on low-rank and sparse matrix decomposition-sparse representation (LRaSMD-SR). First, the relatively pure background is obtained by LRaSMD. Then, the background dictionary model is constructed from the pure background by means of sparse representation. Finally, the reconstruction error is employed to detect the anomalies. The effective experimental tests are conducted using both simulated and real datasets, and the experimental results show that the proposed LRaSMD-SR algorithm possesses a very promising performance of anomaly detection.

The satellite stereoscopic images and the oblique photograph data are used as the data sources. The remote sensing images of Hunan Province obtained from ZY-3 are taken as examples. Based on the tile-pyramid model, the fusion of multiple types of data with different resolutions is realized and a real three dimensional (3D) model with an integrated expression of terrain and texture is constructed. In order to achieve the visualization of data models based on the massive terrain data in 3D scenes, a software named ItelliEarth is developed independently. The research results show that the real 3D model constructed by the proposed method can meet the requirements of relevant specifications, which provides an effective way to realize a large-scale 3D real space model.

The research progress on photodynamic therapy (PDT) in the treatment of gastrointestinal diseases is reviewed, including the related in vitro, in vivo studies and clinical practices. The therapeutic effects and application prospects of PDT for different types of gastrointestinal diseases are explored in order to provide the necessary technical supports for the more applications of PDT in the clinical treatments of gastrointestinal diseases.

Scene understanding is an important research content in information science. Compared with the two-dimensional (2D) data, the three-dimensional (3D) data has many advantages. At present, there are many ways to acquire the point clouds, and meanwhile the point clouds with different acquisition methods have different characteristics. In addition, there lacks a complete and systematic research review on the key techniques for 3D scenes understanding. Thus, the different methods for point cloud acquisition are summarized, and the different point cloud data and related databases are compared and analyzed as well. Based on the current research progress of 3D scene understanding, the techniques for point cloud filtering, feature extraction, point cloud segmentation, and point cloud semantic segmentation in 3D scene understanding are compared and summarized. By the summary of the domestic and foreign literatures published in recent years, the problems occurred in the key technologies for 3D scene understanding are condensed, and the development trend of the 3D scene understanding problems is prospected. The 3D scene understanding based on point clouds is widely used in many fields due to its richness of data. However, as for the scene understanding effect of 3D point clouds, especially the scene understanding of laser point clouds with color information, there are still many contents needed to be investigated in depth.

The recent research achievements on high-power semiconductor lasers in various countries of the world are reviewed. The research progress is mainly introduced in terms of output power, brightness, electro-optical conversion efficiency, beam quality, lifetime, and reliability of GaAs-based near-infrared high-power semiconductor lasers. Combined with the current market analysis, the application prospect of these semiconductor lasers is elaborated. The development trend of high-power semiconductor lasers in the future is forecasted.

The research progress and development status of spectral beam combining (SBC) using high-power fiber lasers is reviewed briefly, and a detailed introduction to the external factors that affect the beam quality is made. The influences of components in the SBC system and the linewidth broadening of the sub-light source array are mainly considered in the evaluation of beam quality. The summary of the existing theories and experimental results as well as the groundbreaking works for improving the beam quality of SBC in Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, is helpful to optimize the design of the built-in parameters of SBC and thus the further development of high-brightness SBC techniques is promoted.

A photoacoustic spectroscopy detection system based on a broadband light source is introduced, and the methane concentration is detected at the optimum modulation frequency of 22 Hz; further, the stability and the minimum detection concentration of the system are analyzed with Allan variance. The research results exhibit that the concentration (volume ratio) detection limit of methane can exceed 1 × 10 -6. The methane concentration is detected by broadband photoacoustic system and direct absorption detection system based on distributed feedback diode laser source with a long path absorption cell simultaneously. The detected results show that the error of methane concentration detection by the broadband photoacoustic system is about 8%.

The type-II spontaneous parametric down conversion is used to prepare the entangled photon pairs. In the Mathematica environment, the properties of frequency entanglement, frequency correlation, quantum interference and the influences of parameters for pulse and crystal on the properties are analyzed. The results show that the continuous optical pumping can be used to obtain the maximum frequency entanglement and interference visibility. When the spectral width of pulsed laser is constant, as the nonlinear crystal thickness increases, the combined two-photon spectra become narrow, the frequency entanglement increases, the indistinguishability decreases, and the interference visibility decreases. In contrast, when the nonlinear crystal thickness is constant, as the spectral width of pulsed laser decreases, the combined two-photon spectra become narrow, the frequency entanglement increases, the indistinguishability increases, and the interference visibility increases. For different parameters of the combined spectral function, a photon pair with frequency anti-correlation, non-correlation and positive correlation can be obtained.

The rapid identification of foodborne pathogenic bacteria is an important task. Compared with the traditional detection methods, Raman spectroscopy is a non-destructive testing method and can simultaneously enhance the identification speed. In order to improve the accuracy and efficiency of Raman spectroscopic identification of Escherichia coil O157∶H7 and Brucella suis vaccine strain S2, a integral classification model is proposed based on the principal component analysis and the Stacking algorithm, whose robustness is improved by the grid search and K-fold cross validation. It is experimentally confirmed that compared with the logistic regression, K nearest neighbor, support vector machine and other single models, the integral model based on the Stacking algorithm possesses the highest accuracy rate of 99.73% the expected result is achieved.