The Hyper-spectral Infrared Atmospheric Sounder( HIRAS) is a new vertical temperature and humidity detection instrument carried by China ’s second-generation polar orbiting meteorological satellite FY-3D. The determination of the zero optical path difference( ZPD) position of in-terferograms is the premise of interferogram overlays. However, in practical applications, it is difficult to determine this position because of the influence of the instruments involved and their environments. In this study, based on measured on-orbit HIRAS data, the influence of ZPD position deviation on data inversion is analyzed. The maximum correlation method and the minimum imaginary part method of complex spectra are used to detect the ZPD position. After the correction, the phase difference between the inner blackbody and deep space is approximately π. The imaginary part of the calibrated spectra is nearly equal to 0, which only represents noise. This method can be well used for HIRAS data preprocessing.

To achieve the accurate detection of 14th-magnitude targets, a large-field-of-view optical detec. tion system was designed. First, according to the selected CCD231-84 E2V photodetector, the initial pa. rameters, such as system entrance pupil and focal length, were calculated. Then, based on the selected pa. rameters, the initial structure was selected, and the form of the Maxutov telescope was selected and im. proved. Then, the detection performance of the design results was analyzed. Finally, a tolerance analysis and optimization of the designed system were conducted so that it could meet the needs of processing and assembly. The design and analysis results indicate that the optical detection system adopts a spherical catadioptric mirror surface, the total length of the system is 350 mm, the full field of view for a 30 μm sur. rounding energy distribution is >86%, and the maximum distortion is <1%. The system has a large field of view, aperture, and relative aperture. It is compact and of suitable size, easy to install, and offers high detection performance, high detection sensitivity, and a wide detection range, making it useful for the ac. curate detection of targets.

Optical losses of silica waveguides are mainly caused by the core sidewall angle and rough. ness of the silica waveguides. In this study, to accurately measure the sidewall angle of an upper-clad. ded waveguide core, we investigate a method using an optical image microscope. Further, we thor. oughly analyze the relationship between the magnification of the microscope objective and the imaging features that cause measurement errors. Then, we compensate the experimental measurement results of the waveguide sidewall angle with the intrinsic systematic errors based on a theoretical analysis, which agree with the values obtained via confocal laser microscopy measurements. Furthermore, by comparing the sidewall angle measurement results of 10 sets of cladded waveguides with the confocal laser microscopy measurements of the waveguides before being cladded, we demonstrate that the measurement error is controlled within ±1°. This provides a feasible method to quickly and accurately measure the sidewall angle of the core layer of cladded silica waveguides, i. e., after deposition of the cladding.

Aiming at the problems inherent to existing targeting algorithms, such as complicated calcula. tions, poor accuracies, and lack of fault tolerances for outliers when solving for the target position using multiple sets of theodolite angle measurement data, a new type of fusion fault-tolerant positioning method based on the shortest distance sum was proposed. By using the sum of the shortest distance method to fuse data from different devices, the fault-tolerant algorithm designed in this study could still obtain the target position, which was mostly unaffected by outliers in the measured data, thereby ensuring the reliability of the positioning results. Simulation results show that the proposed method exhibits a good fault tolerance for the measured data, i. e., when the measurement data are normal, the location result reaches the mini. mum sum of distances, and when the measurement data contain isolated outliers or speckle outliers with a length of less than 3, it can ensure that the positioning results are still reliable and mostly unaffected by out. liers without prior repair. The fault-tolerant location algorithm based on the sum of the minimum distances not only makes full use of the effective measurement data information, but also carries out positioning cal. culations without an outlier detection, ensuring that the positioning results are not distorted. This has im. portant engineering value for real-time and reliable positioning of dynamic targets.

In this study, an iterative phase-shifting algorithm based on the least-squares principle was de. veloped to achieve multi-surface interferometry. The accuracy of interferometry was further improved by incorporating the secondary reflection signal into the least-squares solution equation. First, according to the least-squares principle, the theoretical value was correlated with the actual interference information to derive the least-squares equation. This derived equation was divided into two parts: the iterative calcula. tion of the initial phase and the iteration of the phase-shifting value between two successive frames. From the iterative results of 21 interferograms, an accurate initial phase distribution was obtained. The simula. tion results show that the solution accuracy without considering the secondary reflection signal is 10 nm, and the accuracy improved to 0. 3 nm when the signal is considered in the calculation. In addition, the experimental results of the measured data indicated that the developed algorithm can be used to reconstruct a more accurate phase distribution, and no clutter interference or residual harmonic signal is observed in the results. Therefore, the developed algorithm can be used to achieve high-precision multi-surface measure. ment, which has research value and practical significance.

A back-illuminated p-i-n-i-n separate absorption and multiplication(SAM)solar-blind ultravio. let AlGaN avalanche photodiode(APD)with a low-Al-content p-graded AlxGa1-xN layer and a high/low-Al-content AlGaN multiplication layer was designed. Simultaneously, an AlN/Al0. 64Ga0. 36N distributed Bragg reflector(DBR)structure was inserted to improve the solar-blind photo-response of the designed APD. To fabricate a specially designed AlGaN APD under experimental conditions, the rationality of the device structure needs to be verified in advance. Therefore, the designed APD was simulated using Atlas-Silvaco simulation software, and a conventional APD structure was used as a reference. The simulation re. sults indicate that the designed APD exhibits enhanced optoelectronic characteristics in comparison to a conventional APD, which is attributed to the higher hole-initiated impact ionization coefficient and polariza. tion electric field generated in the same direction as the applied bias field of the designed APD. The de. signed APD exhibits a 10-fold avalanche gain, which is 6. 11×104, and reduced avalanche breakdown voltage compared with a conventional APD. Moreover, the designed APD exhibits good optoelectronic characteristics, and willprovideasolidtheoreticalbasisforthefurtherdevelopmentofAlGaN-based APDs.

Based on cascade Mach-Zehnder modulators(MZMs), a broadband microwave frequency down-conversion method was proposed, and the target distance information was derived from the down-converted intermediate frequency(IF). In the proposed scheme, a linear frequency modulated microwave signal was multiplicated by controlling the bias point and modulation index of MZM. The multiplicated sig. nal was then transferred through an electric time delay to simulate the retuning signal of the target, and the transferring delayed optical sidebands are modulated on MZM. The IF signal carrying the target distance information was obtained after a low-pass filter. Analogous simulations of frequency doubling, quadrupli. cating, and sextupling of central frequencies of 10. 8, 21. 8, and 32. 4 GHz and bandwidths of 2. 7, 5. 6, and 8. 4 GHz are achieved. Moreover, as the electrical time delay increases, the IF linearly decreases. The proposed method is proven to be flexible, highly accurate, and easily implemented, demonstrating its great potential in radar systems for wideband imaging systems.

Two-photon polymerization technology for use in a femtosecond laser was used in the rapid and high-precision processing of a patterned microlens to improve the previous shortcomings, including a complex processing technology, expensive manufacturing technology, and limited pattern design. First, three-dimensional software was used to design the microlens pattern through the deformation of the spherical wave factor, and the two-photon polymerization processing technology for a femtosecond laser was used to process the patterned microlens in the photoresist sample. The sample was then placed in a developer to re. move the unprocessed area and obtain the corresponding patterned microlens. Finally, an imaging test and a light intensity homogenization analysis of the patterned microlens were carried out. An LED light source was placed below the patterned microlens, and the light was successfully focused through the patterned mi. crolens to obtain the corresponding patterns with the same light intensity. The experiment shows that the two-photon polymerization of a femtosecond laser can realize the flexible and controllable processing of a 3D microlens structure, a processing power of 7 mW, an exposure time of 2 ms, a scanning xy-step of 0. 5 μm, and a z-step of0. 8-1. 5 μm, ensuring the smooth surface of the microlens structure and realizing a rap. id microlens processing. The two-photon polymerization technology for a femtosecond laser will play an important role in the processing field such as optical metamaterials, optical microdevices, and integrated optical devices.

High throughput monodisperse nano-liter/pico-liter droplet emulsions are essential in several biomedical applications, such as droplet digital PCR, mass spectrometry, and cell sorting. However, the preparation rate of existing technologies can only reach 1 000/s, which restricts development of the bio. medical field. We present a novel technology for providing monodisperse droplets with high throughput us. ing centrifugal force and a micro-nozzle array. First, we designed the centrifugal droplet microfluidic chip and studied the relationships of droplet diameters with dimensions of the micro-nozzle, centrifugal accelera. tion, and flow rate based on simulation of the droplets’generation using the micro-nozzle;based on these, we obtained optimized design parameters for generating 1-nL droplets. Then, a microfluidic chip was fabri. cated using standard lithography. Finally, high throughput droplet generation was performed. The experi. mental and simulation results indicate that the droplet size is inversely related to the centrifugal acceleration and directly proportional to the micro-nozzle size. The droplet size is stable in a large range of flow rate, which indicates excellent robustness. Homogenous droplets with pre-selectable diameters in a range from 105 to 145 μm were generated with a coefficient of variation of 3% and generation rate of 27 000/s under a cen. trifugalaccelerationrangeof250g to500g, demonstratingultra-high throughputcharacteristicsofthechip.

To further improve the absolute positioning accuracy of a robot, a method for realizing the error prediction based on support vector regression(SVR)was proposed. First, an MDH model was used to es. tablish a kinematic robot model, and SVR was used to establish the prediction model of the rotation angle and position error of a robot. Second, the grid division was controlled based on the spatial accuracy, and the relationship between the sampling points and the calibration accuracy was analyzed to establish an ap. propriate mode for the area division. Finally, the differences between the values of the theoretical and real position coordinates of the robot measured with a laser tracker were used to train the SVR model and com. pensate the single-point position errors. The experimental results indicate that the arithmetic mean error of the robot at the center, and the edge positions, are reduced from 2. 107 mm and 2. 182 mm to 0. 103 mm and 0. 123 mm, respectively. The correctness and effectiveness of the SVR for the absolute positioning er. ror compensation of a robot are also verified.

Aiming at the robot's feedback to the working environment for realizing automatic operation, a piezoelectric six-axis force sensor was designed. Specifically, an eight-fulcrum spoke six-axis force sensor structure combined with the principle of multipoint measurement was proposed herein. Quartz was em. ployed as the piezoelectric material in the force-sensing device. The advantage of such a sensor is its strong decoupling, light weight, large range, and high natural frequency. The parametric modeling of this sensor was carried out using the ANSYS finite element software. The measurement principle of the eight-fulcrum sensor’s feasibility was verified through a six-axis force quasi-static simulation. Moreover, a quasi-static and dynamic calibration experimental platform was built to conduct the calibration experiment. The experi. mental results show that for the proposed sensor, the nonlinearity is less than 0. 5%, inter-dimensional in. terference is more than 4%, and natural frequency is over 6 kHz. From the theoretical analysis and experi. mental results, we confirmed that the sensor structure could complete the six-axis force measurement and that the coupling between the dimensions was eliminated. The demand for dynamic measurement was thus fulfilled. The proposed sensor’s structure provides the theoretical and design bases for the design of light weight large range sensors, which is of important reference significance for the design and development of other types of piezoelectric six-axis force sensors.

A novel deep instance feature chain learning network for panoptic segmentation(INFNet)was developed to solve the problem of failure of target boundary segmentation caused by insufficient instant fea. ture extraction in panoptic segmentation. This network consisted of a basic chain unit, whose functions were divided into two types, feature holding chain and feature enhancement chain, based on the different methods of processing feature information by the unit structure. The feature-holding chain represented the input stage of the extraction of a chain network feature, in which the integrity of the input information was guaranteed, and then this feature was transmitted to the feature-enhancement chain structure. The feature-enhancement chain increased the network depth and improved the feature extraction ability through its ex. tension. INFNet could obtain adequate edge feature information and improve segmentation accuracy, ow. ing to the robust depth-stacking characteristics. The experiment results for the MS COCO and Cityscapes datasets showed that our INFNet was superior to similar existing methods in terms of segmentation accura. cy. Compared to the Mask RCNN instance segmentation structure widely used in panoptic segmentation networks, the segmentation accuracy of INFNet increased by up to 0. 94%.

A photoelectric detection system based on a lock-in amplifier was developed. The noise charac. teristics of each part of the system, such as the photomultiplier, amplifier, and high-voltage power supply, were analyzed quantitatively. Research has shown that the ripple of the high-voltage power supply has a significant effect on system noise. Three high-voltage power supplies with different ripples were installed in the system, and the noise was analyzed. The results showed that in darkness, system noise was mainly contributed by photomultiplier anode dark noise and noise from the high-voltage power supply ripple linear. ly coupling into the system. In light, photomultiplier anode shot noise contributed the most to the system noise, and noise contributed by the high-voltage power supply ripple showed a positive correlation with both light current and ripple value. Based on the above findings, the transfer function of the high-voltage power supply was analyzed, and its feedback coefficient was optimized. A high-voltage power supply with high stability and low ripple(less than 5 mV)was designed. The signal-to-noise ratio test of the system showed that the signal-to-noise ratio could be significantly improved by applying self-developed high-voltage power supplies, which were 38% and 125% higher than those of other high-voltage power sources(ripples of 15 and 50 mV, respectively).

A froth-flow feature detection method based on an improved ORB in the NSST domain was de. veloped and applied to flotation dosing state recognition to solve the problems of continuous movement, light effects, and noise interference of flotation surface images, which lead to difficulties in flow feature de-tection. First, two adjacent froth images were decomposed through NSST. Multiscale high-frequency sub-bands were denoised using a scale correlation coefficient and then divided into multiple inner and outer lay. ers. The points of interest were subsequently extracted through modulus maxima detection in each inner layer, and the feature points were extracted through non-maximum suppression between the upper and lower layers. Second, a multiscale BRIEF descriptor was adopted to describe these feature points, the search matching area was dynamically adjusted according to the movement trend of the bubbles. The froth-flow features were then calculated based on the matching results. Finally, a line-and-column autoencoder ex. treme learning machine was constructed to fuse the foam shape, size distribution, and flow features, and the dosing state was recognized by the adaptive random forest method. The experimental results showed that the improved ORB was slightly affected by noise and illumination. The flow feature detection efficien. cy and the detection accuracy were significantly better than those of existing methods. The proposed meth. od could characterize the flow characteristics of the froth surface accurately in different dosing states. The average accuracy of dosing state recognition reached 97. 85%, which was significantly higher than those of existing methods. This study lays a foundation for future research on dosing quantity optimization control.

A pedestrian detection network based on the weight learning of fusing multimodal information was developed to address the issues of the pedestrian detection method based on infrared and visible modal fusion in adapting to changes in the external environment. First, unlike the fusion method used in several recent studies in which two modalities are stacked directly, the weight learning fusion network reflects dif. ferent contributions of the modalities to the pedestrian detection task under different environmental condi. tions. The differences between the two modalities were determined through dual-stream interaction learn. ing. Next, based on the current characteristics of each modal feature, the weight learning fusion network assigned the corresponding weights to each modal feature to generate the fusion feature by performing weighted fusion autonomously. Finally, a new feature pyramid based on the fusion feature was generated, and previous information about the pedestrian was improved by changing the size and density of prior boxes to complete the pedestrian detection task. The experimental results indicated that the log-average miss rate of the Kaist multispectral pedestrian detection dataset reached 26. 96%, which was 2. 77% and 27. 84% lower than that of the direct stacking method and baseline method, respectively. The adaptive weight fu. sion of infrared and visible modal information could effectively be used to obtain complementary modal in. formation to adapt to external environmental changes and significantly improve pedestrian detection perfor. mance.

To address the needs of the Internet of Vehicles, a plug and play visible light signal transmis. sion control device was designed for a vehicle lamp. In this device, the signal was first transmitted to a light-emitting diode(LED)driver via the DC power line and output as flashes of the vehicle LED lamp. The device was composed of a front-end circuit and a back-end circuit. The power switching was con. trolled by a high-speed on-off circuit in the front-end circuit based on the received input signal. The signal in the DC power line was then transmitted to the back-end circuit. Based on the signal, the VLC signal was sent by driving the flashing LED lamp. A capacitor was selected as the energy storage module after charge – discharge analysis. The communication between the front-end and back-end circuitry was achieved with 2-frequency shift keying(FSK)modulation and Modbus protocol. The device tests showed that the signal could be transmitted over the DC power line at minimum speeds of 4 Mb/s. The design and testing of the VLC parking lock control module also proved the feasibility of the communication and control functions. Thus, the proposed device could be a reference design for DC power line communication for an LED VLC system.

In image feature extraction, sample labels are rarely completely true and effective. This often leads to a significant decrease in the accuracy of an image classification framework. In addition, existing label recovery algorithms often must deal with a bottleneck problem in which noisy samples are difficult to reuse. Therefore, this paper proposes a subset-divided iterative projection bagging algorithm for noisy-label recovery. First, the proposed algorithm extracts small-scale subset information randomly and re. peatedly. It then integrates principal component analysis, neighbor graph regularization, K-nearest neigh. bor, and other techniques to achieve effective dimension reduction and iterative projection integration of sample images. Finally, class-label recovery is conducted by implementing the majority voting principle. This study uses common databases as experimental objects and conducts several comparisons and analy. ses of various recovery algorithms using different indicators. Experimental results show that the proposed algorithm effectively corrects the noisy labels of samples, and the classification accuracy of the default framework is improved by as much as 16. 9% and 8. 1% for the Yale B and AR databases, respective. ly. Compared with the state-of-the-art algorithm, the classification accuracy of the proposed algorithm is improved by 4. 3-4. 7%. The proposed algorithm also has good scalability and can ensure the integrity of sample data.

Craniofacial reconstruction is used to estimate facial data for a given unknown skull data, and it has been widely applied in various fields such as forensic science, anthropology, and criminal investiga. tion. It is time-consuming and laborious to extract feature points manually during craniofacial reconstruc. tion. In this study, radial curves uniformly distributed on the face were automatically extracted starting from the nose tip as a feature representation of the three-dimensional craniofacial model based on its geome. try structure. The extracted radial curves and skull data were used as training sample data to establish a sta. tistical model for craniofacial reconstruction. Previous knowledge obtained using the statistical model and skull data were used to estimate the data for the face. The experimental results showed that when the craniofacial statistical method based on the radial curves was applied, the craniofacial reconstruction accuracy improved by 2. 95 times;moreover, the reconstructed speed was 4. 01 times faster than that of the cranio. facial reconstruction method based on principal component analysis. Therefore, our method reduces the di. mension of craniofacial data and improves the accuracy and speed of the reconstruction results.

For the existing binocular position and attitude measurement system with only two target points, it is impossible to use additional calibration objects to calibrate the two cameras without a common field of view. In this study, a calibration method for a binocular camera without changing the system struc. ture and instead directly using the existing equipment was proposed. First, the relative position between the two camera coordinate systems and the common coordinate system was fixed, and multiple sets of co. ordinate data were acquired by moving the calibrator several times. An improved algorithm based on a di. rect linear transformation(DLT)method was used to optimize the relative transformation from each cam. era coordinate system to the common coordinate system. Finally, the relative conversion between the two camera coordinate systems was obtained using the optimized relative conversion from each camera coordi. nate system to the common coordinate system. The distance between the two target points could be ob. tained using the relative conversion between the two camera coordinate systems, and this distance was se. lected as the evaluation index of the calibration accuracy. The experimental results show that the mean square error of the distance between the target points obtained using this method is less than 0. 2 mm, which meets the practical application requirements.

Due to its small unit size, dense array, weak fluorescence signal, and vulnerability to light dis. tribution, it is difficult to locate the sample points of honeycomb stacked digital PCR microarray images. In this paper, an algorithm of three-channel honeycomb fluorescence spots addressing based on morpholo. gy, and a digital PCR image information extraction method are proposed, which can quickly and effective. ly identify the weak fluorescence information of biomolecules in microarray chips. The image registration fusion is carried out for different channels to make the sample points orderly. Select effective spots area by enhancing the contrast of the image. Use improved gamma algorithm to remove uneven distribution of illu. mination. Identify, locate and segment the closely arranged spots on the microarray chip based on morpho. logical algorithm, and then, extract the fluorescence information of biomolecules of each spot. Using this method, the image processing time of a digital PCR chip with about 20 000 microcells is less than 20 s. Compared with the existing spot addressing methods, the processing time of the same number of samples can be reduced by three orders of magnitude. Compared with the results of standard instruments, the accu. racy of sample recognition is 98. 79%, and the accuracy of biological information calculation(copy num. ber)is 96. 2%. In this paper, the three-way honeycomb fluorescence spot addressing algorithm and digital PCR image information extraction method based on morphology are proposed to overcome the difficulty of locating the spots in the honeycomb stacked microarray fluorescence picture. Compared with the existing methods, biological information can be obtained quickly and accurately, which lays a foundation for pre. cise quantification of digital PCR technology.