Snapshot compression imaging technology can obtain the three-dimensional spatial-spectral information from the target within a single exposure imaging, which has a significant advantage in the detection and identification for moving targets compared with the traditional scanning imaging method. With the development of information technology and computer processing performance, computational imaging has gradually become one of the most important technologies for solving optical imaging problems. By building a physical model of the imaging device and mathematically optimizing the back-end processing, which can break through the limitations of the imaging model and detector, the traditional two-dimensional imaging can be extended to more observation dimensions. In this paper, the current research status of snapshot hyperspectral imaging technology based on coding compression is summarized from three aspects: spatial coding, wavelength coding and phase coding. And we summarize and analyze the development trend of traditional methods and deep learning methods, as well as look forward to the development of snapshot hyperspectral imaging technology based on coding compression.

The article centers around conducting simulation analysis and experimental research on thermal analysis and management of air-cooled fiber lasers. It addresses heat transfer issues concerning gain fibers and optical modules within these lasers, establishing respective theoretical models and validating them through experiments. Furthermore, it proposes a solution tailored for portable air-cooled fiber lasers to manage thermal issues at the kilowatt level, focusing on optimizing system parameters, dispersing waste heat effectively, and minimizing overall thermal resistance.

The temperature of the laser crystal in a solid-state laser can affect the performance of the laser, such as power stability and spot quality. A high-precision temperature control system has been designed to meet the temperature control requirements of crystals in solid-state lasers. The system adopts a non-equilibrium bridge combined with thermal sensors for temperature sampling, a second-order active low-pass filter to filter out high-frequency noise interference, and a high-precision ADC scheme to improve sampling accuracy. Two half bridge driving chips form a semiconductor cooler (TEC) driving circuit, which has strong driving ability. By improving the PID algorithm, the response speed is increased, and the working current of TEC is adjusted more quickly and accurately to achieve highprecision and stable temperature control effects. The GD32F103RCT6 microcontroller is used as the control core to build a temperature control system. The experimental results show that the system can achieve stable control within the set temperature range of ± 0.005 ℃.

To address the challenges of observing colorless and transparent samples with traditional microscopy techniques, a phase microscopy imaging system based on the Transport of Intensity Equation was designed. This system utilizes the axial differential approximation method and the in-focus image approximation method, combined with a binocular CMOS camera to simultaneously collect overfocused and underfocused images, thereby solving the Transport of Intensity Equation and achieving phase retrieval of the sample. The system also integrates an autofocus algorithm and a phase-related field of view correction algorithm to reduce the impact of human and equipment errors. Tests have shown that the system achieves a resolution of 2.19 m under a 10 objective lens. In practical applications, it has not only successfully performed phase retrieval on human blood smears, observing the unique structure of red blood cells, but it can also be used for scratch detection in petri dishes. Due to its low cost, ease of implementation, and broad application potential, the system demonstrates significant advantages in observing colorless and transparent samples.

A novel Fabry-Perot interferometer (FPI) pressure probe was prepared using ultraviolet (UV) adhesive and ceramic ferrule for high-sensitivity pressure measurement. First, a layer of UV glue is evenly coated on the end face of the optical fiber ceramic ferrule, then a section of flat single-mode fiber (SMF) is inserted into the optical fiber ceramic ferrule, adjusting the distance between the optical fiber end face and the ceramic ferrule end face UV glue film, and an FPI with excellent performance is made. Due to the excellent elastic deformation of UV adhesive, FPI is sensitive to changes in air pressure. The experimental results show that the pressure sensitivity and temperature sensitivity of the sensor are -10.712 nm/MPa and -1.449 nm/℃, respectively. In order to compensate for the impact of temperature changes on sensor pressure measurement, a temperature sensitive and pressure insensitive fiber Bragg grating (FBG) was cascaded with FPI to eliminate the influence of temperature on FPI pressure measurement. This sensor has the advantages of simple structure, convenient preparation, low cost, high sensitivity, good repeatability and stability, and has certain application prospects in industrial production.

In the narrow pulse laser driving circuit, the transmitting and receiving antennas operate in dual polarization mode, which is affected by non ideal factors, making it easy for the polarization direction of the source signal to partially transfer to other polarization directions during signal transmission, resulting in polarization interference and introducing unexpected interference signals, reducing the stability of the control process. Therefore, a delay feedback control method for narrow pulse laser driving circuit is proposed. By describing the ratio of signal power to cross polarization interference through cross polarization isolation, quantifying the interference size, using blind source separation to restore the source signal, using kurtosis function to determine the optimal step size, extracting non zero kurtosis signals from multiple independent components as source signals, and achieving interference signal suppression. Construct a discrete orthogonal virtual circuit in the driving circuit of a narrow pulse laser, calculate the virtual components of the actual voltage and current orthogonal and unfold feedback to achieve delayed feedback control of the narrow pulse laser driving circuit. The experimental results show that the proposed method has good stability and can obtain more satisfactory delay feedback control results for narrow pulse laser driving circuits.

In the process of natural gas transmission and oil extraction, in order to ensure the safety of personnel and equipment need to accurately detect CH4 gas concentration, based on the TDLAS laser gas detection technology designed a data acquisition system capable of high-precision measurements of CH4 gas concentration, the system selects the STM32H743 as the main control chip, adopts the ADC's dual mode, utilizes the internal ADC1 collects the fundamental and second harmonic signals, and the internal ADC2 collects the temperature and pressure data of the gas, and then the data are averaged, and the averaged calculation parameters are controlled by the master chip, and the data are transmitted and stored through DMA, and finally the master chip integrates and processes the multiple sets of data and sends them to the upper computer through the virtual USB interface in accordance with the specified data protocol. After actual testing, the system can collect and process data in a high-speed and stable way, and can be applied to other gas TDLAS detection systems.

Remote sensing images object detection is used in mineral exploration, transportation, national defense and military, emergency rescue and disaster relief and other fields. However, the common rotation object detection algorithm model applicable to remote sensing images is too large, difficult to deploy and can not meet the requirements of real-time detection, ignoring the balance between accuracy and speed. To solve the above problems, a rotation object detection algorithm YOLOV5s-R based on YOLOv5s is proposed. Firstly, angle parameters were added on the basis of YOLOv5s, then the horizontal bounding box was modified into oriented bounding box to adapt to the angular diversity of remote sensing image objects. Secondly, the Circular Smooth Label was introduced to avoid the angle mutation problem caused by the periodicity of angle regression. Then, the Efficient Channel Attention module was introduced to improve the ability of the model to extract important features. Finally, the Adaptively Spatial Feature Fusion module was introduced to solve the inconsistencies between different feature scales inside the feature pyramid. On the dataset DOTA, the experimental results show that the mAP50 reaches 75.6%, the mAP50∶95 reaches 46.7%, and the FPS reaches 81.9. Compared with the base model, mAP50 and mAP50∶95 increased by 1% and 3.1% respectively, and FPS increased by 85.9%. Therefore, YOLOv5s-R achieves more accurate and high-speed remote sensing images detection, achieving a good balance between accuracy and speed.

Aiming at the problems such as slow detection speed, unstable accuracy and great destructiveness of traditional crucible thickness measurement methods, a deep learning based crucible thickness measurement algorithm was proposed. Firstly, a set of visual hardware platform was built, and the image acquisition system of crucible thickness was constructed. The images are then captured by an electron microscope carried by the robotic arm. Secondly, by calculating Brenner function value and drawing corresponding curve, using wavelet transform peak detection algorithm to calculate the peak value of Brenner function curve and determine the internal focus image. The convolutional neural network is used to extract the features of spots, realize spot recognition and detection, and count the number of spots to get a bimodal curve. The peak value of the spot curve is calculated by wavelet transform peak detection algorithm to determine the external focusing image; Finally, the crucible thickness was obtained by averaging the frame difference of the two internal and external focusing images and the moving velocity of the electron microscope. The experimental results show that the measurement error of this method is less than 1 mm. The proposed method can accurately measure the thickness of quartz crucible and has high reference value for crucible quality detection.

Line-structured light scanning measurement technology based on galvanometer mirrors has wide applications due to its outstanding accuracy and high efficiency in the field of 3D measurement. However, the difficulty in precisely aligning the laser with the galvanometer's rotation axis during installation significantly impacts the calibration accuracy of the galvanometer coordinate axis, thereby affecting the overall measurement system's accuracy. Consequently, the precise calibration of the scanning galvanometer coordinate axis has become one of the key challenges in achieving high-precision measurements. A nonlinear algorithm is proposed which uses iteratively approximating to find the optimal solution based on the nonlinear characteristics of the equations to be solved, and the precision calibration of the galvanometer coordinate axis is realized. The theory and processing steps of this calibration method are introduced, and by conducting multiple measurements on a triangular calibration block, it is found that the absolute error of this method ranges from 0.201 1 mm to 0.492 7 mm, lower than the values of 0.313 5 mm to 0.770 3 mm using the traditional least squares method, which indicates that the proposed method significantly improves both accuracy and stability. Therefore, the method provides a new solution for high-precision measurement based on scanning galvanometers, and presents substantial application values.

To solve the problem that traditional Canny algorithm is not accurate to detect the edge of a drop, which leads to a large contact angle measurement error, this paper proposes an improved Canny-Devernay subpixel edge detection algorithm. The algorithm improves traditional Canny algorithm by replaceing gaussian filter with fast-guided filter, calculateing gradient and direction through the operator based on PLIP model, and adopting a local maximum inter-class variance method to achieve adaptive threshold selection. Finally, the improved Canny algorithm is combined with Devernay correction algorithm to realize sub-pixel edge detection. Experiments show that compared with the traditional Canny algorithm, the improved algorithm has more advantages in retaining droplet edge information. When the contact angle measurement method based on this algorithm is detected with a contact angle standard chip, the absolute error and mean of variance can reach as low as 0.011° and 0.009°.

Aiming at the problem of low detection accuracy of small targets in remote sensing images due to complex background, small size and dense arrangement, a remote sensing small target detection method integrating receptive field amplification and feature enhancement is proposed. Using YOLOv8s as the baseline network, the method firstly constructs a receptive field amplification module for the feature extraction part of the backbone network, and efficiently captures the global feature information through the Bi-Level Routing Attention (BRA); secondly, it constructs a shallow feature fusion structure in the feature pyramid part, and adds the improved coordinate spatial attention (CSA) in the transverse connectivity part of the shallow feature map, in order to enhance the feature information of the small targets; Finally, the detection results are post-processed by an improved non-maximum suppression (NMS) algorithm to adapt to the detection of objects with different densities Experiments are carried out on the DIOR remote sensing image dataset, the mean average precision (mAP) reaches 90.3% when the intersection and concurrency ratio threshold (IoU) between the predicted frame and the real frame is 0.5, which is 3% higher than that of the original model; and the mAP reaches 71.3% when the IoU is 0.5∶0.95, which is 6.1% higher than that of the original model, and the experimental results show that the improved model has a good application value for the small target detection in remote sensing images.

For non-contact target sign detection by millimeter wave radar, a fast-moving multi-target scene was studied through simulation experiments. First, the measurement information of the target position and speed is obtained, and the corresponding relationship between each measurement distance and angle is obtained through an innovative algorithm, and then the joint probabilistic data association (JPDA) and nearest neighbor data association (NNDA) are used respectively. Multiple targets are tracked, and then their sign phase information is extracted, and the short-time Fourier transform algorithm is used to estimate the frequency of the processed sign signal. Through experiments, it can be seen that for moving targets with speeds up to 0.9 m/s, this detection method can obtain the target's breathing frequency with a small error; however, because the heartbeat signal is relatively weak, it can only be detected in a static state. Compared with other radar sign detection methods, this method can detect the breathing frequency of the speed of 0.9 m/s at a root mean square error detection rate of 0.035, and can detect multiple targets simultaneously without being easily confused.

Aiming at the problem of poor model generalization performance caused by the scarcity of labeled samples in the target domain, a few-shot hyperspectral image classification method based on cross-domain mixup and self-supervised learning(FSC-CMS) is proposed. First, few-shot learning is used to extract meta-knowledge from the source domain that is more beneficial to target domain classification. Secondly, apply Mixup technology to few-shot learning, perform feature-level Mixup on the query sets of the source and target domains, expand the distribution of the target domain data through the source domain data, increase the diversity of the target domain data, and thus improve the generalization performance of the model. Finally, the few-shot learning process is constrained through selfsupervised learning in the target domain to obtain a more robust feature representation, thereby alleviating the over-fitting problem of the model. A large number of experiments were conducted on two public hyperspectral datasets. Compared with existing mainstream methods, the average accuracy of the proposed method increased by more than 3.2% and 3.6% respectively.

Aiming at the problem that detailed information features are easy to be lost in the fusion process of infrared and visible image fusion algorithm, this paper proposes an infrared and visible image fusion algorithm based on shuffle attention mechanism and residual dense network. Firstly, the encoding network downsamples the source image at different scales to obtain the feature map with rich semantic information. Then the shuffle attention residual fusion network fuses the feature map extracted from the encoding network, and the shuffle attention mechanism aggregates the feature maps through the channel attention and spatial attention shuffling, and utilizes the residual dense connection to maximize the retention of effective image information on the aggregated feature maps. Finally, the decoding network reconstructs the image map through up-sampling. Compared to other fusion algorithms, the fusion images produced by the algorithm proposed in this paper demonstrate a clear advantage in terms of clarity, especially when dealing with complex situations such as blur, occlusion, and smoke, as evidenced by subjective evaluations. This suggests that the algorithm may have a competitive edge in the field of image fusion, particularly in generating clearer fusion results when handling images under complex conditions. In the objective index comparison, the fused images of the algorithm proposed in this paper have different degrees of improvement and achieve the optimal value in the index criteria of entropy, mutual information and peak signal-to-noise ration, which are 6.930, 13.860, 17.144 and 0.574, respectively.

Aiming at the problems that remote sensing image features are difficult to extract and the existing image registration framework has low registration accuracy and efficiency, a bidirectional remote sensing image registration method that combines multi-order features and cross-spatial attention is proposed. First, cross-spatial attention is designed to retain multi-scale accurate spatial structure information into channels, and embed it into efficient network blocks to focus on capturing the key information of the image. Secondly, a multi-order feature adaptive fusion module is proposed to be used in feature extraction to adaptively fuse low-order and high-order features to improve the accuracy of registration. Finally, an enhanced feature matching method is designed to analyze the similarity of features more accurately, establish a two-way matching relationship, and use secondary affine transformation to improve the accuracy and reliability of registration. This method achieved 94.0% correct keypoint probability (PCK) on the Aerial Image data set when =0.05 (: normalized distance threshold), and the average registration time reached 0.93 seconds. The results show that this method significantly improves the registration accuracy and efficiency of multi-source heterogeneous remote sensing images.

Multi perspective registration can maintain the geometric consistency of objects in the scene, but there are many occlusion situations, which limit visibility and integrity under different perspectives. In this regard, a pulse coupled neural network based multi perspective laser image point cloud registration method is proposed. By analyzing the pixel noise response and grayscale distribution characteristics of multi view laser images, the key parameters of each neuron in the pulse coupled neural network are obtained, and the dynamic threshold corresponding to the neuron is determined to achieve multi view segmentation of laser images. Calculate the 3D feature descriptors of each point in the multi view laser image point cloud separately, perform nearest neighbor relationship matching, construct a point cloud relationship set, identify erroneous relationship points through triplet constraint optimization relationship set, and construct an objective function based on the sum of squared errors between matching point pairs in the relationship set. By optimizing the objective function, determine the optimal multi view laser image point cloud registration scheme. The experimental results show that after the application of the proposed method, the internal uniformity, regional contrast, and maximum Shannon entropy of the region are larger, the point cloud overlap and false matching relationships are less, and the Q value is reduced. It can effectively improve the accuracy of point cloud registration results for multi view laser images.

In order to improve the resolution of low light lidar images, a low light lidar image super-resolution enhancement technique based on bilateral filtering is proposed. The Gaussian weighted bilateral filtering method is used to calculate the spatial proximity between pixels and the similarity coefficient of illumination intensity, accurately adjusting the brightness difference between pixels to generate a preliminary low illumination estimation image. The guided enhancement method is used to suppress noise, and the logarithmic transformation method is combined to improve the overall brightness and contrast of the image. The super-resolution of the low illumination lidar image is enhanced by calculating the weights between adjacent pixels. The experimental results show that after applying this technology, the brightness value of low light lidar images remains stable at about 130, the clarity is improved to about 21, the standard deviation exceeds 70, and the information entropy reaches over 7.6 bits, significantly improving the quality of lidar images under low light conditions.

In the process of extracting image contour features, interference from noise can lead to unclear dependency relationships between features, which affects the accuracy of feature information extraction results. Therefore, a laser image contour feature extraction method based on conditional generative adversarial networks is proposed. Firstly, the two-dimensional Otsu function is selected as the adaptability evaluation index of the bee colony algorithm, and optimization is carried out for the initialization population and bee search strategy; Then, using the sine and cosine method and the improved bee colony algorithm, the optimal segmentation threshold of the laser image is obtained by searching for the global optimal solution; Finally, in order to capture the global dependency relationship between features, residual structures and layered dilated convolution modules are integrated into the conditional generative adversarial network, combined with cross attention modules to ensure the smoothness of laser image contour lines. Meanwhile, by utilizing spectral normalization techniques and Leaky activation functions, the training process of the model is effectively stabilized, improving the comprehensiveness and accuracy of laser image contour feature extraction. The experimental results show that this method can obtain high-precision contour feature extraction results from laser images.

The current low resolution laser image reconstruction effect is not ideal, resulting in low reconstruction accuracy and long reconstruction time. In order to overcome the shortcomings of current low resolution laser image reconstruction and improve the accuracy of low resolution laser image reconstruction, a low resolution laser image highresolution reconstruction method based on visual communication is proposed. Firstly, the low resolution laser image to be reconstructed is collected, and the wavelet transform algorithm is used to preprocess the original low resolution laser image to eliminate noise interference. Then, Retinex theory is used to reconstruct the preprocessed laser image, and the histogram equalization method of visual communication technology is introduced to improve the contrast of the laser image. Finally, a low resolution laser image reconstruction simulation experiment is conducted, and the experimental results show that the method proposed in this paper effectively solves the problem of low resolution laser image reconstruction. The clarity of the laser image is significantly improved, the structural similarity exceeds 0.97, the peak signal-to-noise ratio reaches 28 dB or more, and the laser image reconstruction time is controlled within 10 s. Moreover, the overall performance of the reconstruction is far superior to other low resolution laser image reconstruction methods, and it has higher practical applications value.

In the process of image processing in frequency-domain optical coherence tomography, recursive neural networks are mainly used to achieve feature fusion. During the operation process, there is a gradient vanishing situation, which leads to low MAP (average accuracy) of multi-scale feature fusion results. Therefore, a multi-scale feature fusion method for frequency domain optical coherence tomography images based on CNN (Convolutional Neural Network) is proposed. Establish a network architecture for denoising scanned images based on generative adversarial networks, and generate high-quality scanned images without noise information through the original image domain. Using the principle of discrete wavelet transform, the denoised image is decomposed into multiple sub images. By constructing a grayscale gradient co-occurrence matrix, multi-scale image feature vectors are extracted. Starting from the local and global contrast of the image, calculate the image adaptive adjustment coefficient to achieve image detail feature enhancement processing. Finally, a feature fusion model is constructed based on convolutional neural networks, and multi-scale feature fusion results are obtained through matching analysis and concatenation processing of enhanced features. The experimental results show that after the application of the new research method, the MAP value of the multi-scale feature fusion results of frequency domain optical coherence tomography images is higher than 0.8, proving that it can effectively fuse features of different scales.

With the rapid development of satellite technology and rocket carrying capacity, LEO(Low Earth Orbit) satellite broadband communication system equipped with inter-satellite laser communication technology has become a new trend. In order to maintain the accurate initial alignment and establish the link for inter-satellite laser communication, ensure the stable signal transmission, reduce signal loss and interference, and improve the quality and reliability of communication, the process of initial alignment for inter-satellite laser communication is analyzed, and a new ephemeris format and ephemeris broadcast format under the overall consideration of satellite resources is proposed in this paper. A highly reliable communication bus has been designed according to the communication requirements, giving full consideration to various working conditions in the process of inter-satellite laser link establishment process, which has greatly improved its achievement ratio and reliability. The results of on-orbit application prove that the design scheme is reasonable and feasible, which can guide the laser communication terminal to complete the initial alignment efficiently and rapidly, and realize the long-term stable chain building with the aid of ephemeris data, and has great significance for the future research and application of LEO satellite communication network.

In order to solve the problems of low efficiency, slow convergence and low accuracy of traditional registration methods, an improved ICP point cloud registration method considering the key points of multi-geometric feature information classification center was proposed. Firstly, the point clouds are downsampled by the voxel grid, and the ground points are removed to speed up the registration efficiency, and then the covariance is used to solve the eigenvalues for the geometric feature information analysis of the non-ground point clouds, and then the Euclidean distance method is used for clustering. The center point of the cluster was extracted as the key point, and the feature key points were described by the FPFH algorithm, so that the center key points of the feature with the same name were correctly paired, and the initial transformation estimation matrix was obtained. Finally, the bidirectional KD-tree and point-to-area ICP algorithm with improved nearest neighbor distance ratio are used for accurate registration, and the Tukey loss function is introduced to resist outlier noise. Compared with the four methods, the results show that the RMSE of the proposed algorithm is 0.202 6 m, which takes 19.426 seconds, and the registration accuracy and efficiency are higher.

Firstly, the diffraction field distribution of a two-dimensional sinusoidal grating is derived through theoretical deduction, and based on the Fraunhofer diffraction theory, the expression of the field distribution in the target plane for a Laguerre-Gaussian beam passing through the grating is obtained. The numerical simulation results show that when a Laguerre-Gaussian beam passes through a two-dimensional phase sinusoidal grating generated by orthogonal overlapping of two one-dimensional gratings with same frequency and period, an N*N array of vortex beams consisting of the same topological charge is generated at the target plane. However, the intensities of vortex beams with different diffraction orders are unequal. To overcome this drawback, the grating is optimized using the Simulated Annealing algorithm. A 3*3 array with optical intensity inhomogeneity of 0.11‱ and diffraction efficiency of 81%, a 4*4 array with optical intensity inhomogeneity of 0.03‱ and diffraction efficiency of 72%, and a 5*5 array with optical intensity inhomogeneity of 5% and diffraction efficiency of 71% are obtained. The results demonstrate the simplicity and practicality of this method for generating vortex beam array, which will provide technical support for the application of vortex beam array with the same topological charge.

To meet the challenge that the current semantic segmentation methods of LIDAR point cloud image are difficult to balance the processing speed and accuracy, and takes up huge computing resources, a lightweight algorithm based on codec structure is studied. This study firstly uses spherical projection to project 3D point clouds to two-dimensional planes, then designs a lightweight two-dimensional semantic segmentation network MobileSeg based on MobileNetV3 to segment projection map, and finally projects the segmentation results back to the 3D point cloud. The algorithm reduces the occupation of computing resources by reducing the dimensionality of point cloud images and lightweight backbone networks, and avoids various difficulties faced when using neural networks to directly extract features from sparse point clouds. The average intersectional union ratio (mIOU) of the algorithm on the SemanticKITTI dataset is 51.7%, and the parameters of the MobileSeg-S and MobileSeg-L models are 0.9M and 3.2 M, respectively. The accurate and lightweight point cloud semantic segmentation method has a certain application prospect in the field of automatic driving.

The time-varying fiber optic grating sensing signal is usually affected by noise, environmental interference, etc. In order to improve signal quality, a precise subdivision method for time-varying fiber optic grating sensing signal under FPGA technology is proposed. Analyze the principle of fiber optic grating displacement measurement and the causes of moire fringes, and use them as the basis for moire signal subdivision; By combining hardware and software subdivision methods using FPGA technology, a time-varying fiber Bragg grating sensing signal subdivision circuit is designed. Based on the subdivision circuit, a spatiotemporal conversion subdivision method is introduced to perform signal subdivision processing, in order to improve signal subdivision accuracy. The experimental results show that the proposed method has a maximum subdivision error of only 0.58 mm at different subdivision multiples, and a minimum subdivision error of only 0.19 mm in noisy environments, indicating that the method can achieve high-precision subdivision of time-varying fiber optic grating sensing signals.

Compressed sensing is gradually applied to optical imaging systems because it only needs a very small number of observations to reconstruct the original signal. Aiming to address the problem that the existing reconstruction algorithms have low reconstruction accuracy in video imaging systems with limited sampling resources, a phased reconstruction algorithm is proposed. In the first stage, the algorithm uses the reconstructed keyframes as reference frames to reconstruct the non-key frames through motion compensation and residual reconstruction. After this, it groups the reconstructed non-key frames, enhances them further, and outputs them. In the second stage, the algorithm dynamically selects multi-hypothesis matching blocks from the current frame and the keyframes before and after for the non-key frames reconstructed in the first stage. It then establishes a residual sparse model and completes the reconstruction to output the video reconstruction results. The experimental results show that the average PSNR value of this algorithm is above 41.5 dB, and the average SSIM value is above 0.97. These values represent significant improvements compared to several existing excellent reconstruction algorithms. Compared with the most representative multi-hypothesis algorithm, the average PSNR value and average SSIM value of this algorithm are improved by about 9.1% and 17.3% respectively, leading to better video reconstruction quality.

In the last few years, phase shifting profilometry has gradually applied High speed real-time three-dimensional measurement for dynamic scenes. The motion of an object in multiple phase shifted fringes not only causes position mismatch, but also introduces additional phase shift, making traditional Phase Shifting Profilometry unable to accurately measure moving objects. This article centering on moving objects, proposes a two-step high frame rate 3D reconstruction method utilizing dual sampling. Sampling two moving object stripe images within one stripe projection cycle, analyzing the impact of object motion on phase under the same projected stripes and between different projected stripes, establishing stripe description equations for different situations, and finally achieving three-dimensional reconstruction of moving objects based on two stripe images with known ambient light. The experimental results show that this method not only achieves three-dimensional measurement of moving objects, but also effectively improves the reconstruction frame rate.

Fiber-optic current transformer (FOCT) is an important primary equipment in UHVDC transmission system. FOCT based on PZT modulation is a widely used modulation mode, in which the polarization-maintaining fiber delay ring is the most sensitive to vibration, that is one of the main errors of FOCT under vibration. In this paper, the influence of vibration on FOCT error is analyzed on mechanical and optical characteristics. It is pointed out that the external vibration will produce an additional modulation superimposed on PZT modulation, which will change the harmonic ratio of the output signal and introducing an error. Both theoretical analysis and experimental results show that the mechanical resonance characteristics of the fiber delay ring have a significant influence on the error. The resonance curve and calculation formula for multiple resonant points are obtained, and the error for different frequencies, different accelerations and different measured currents are given. The theoretical analysis is basically consistent with the experiment.

The meridian design is one of the keys for the ophthalmic progressive addition lens design. A new principle meridian model, giving out the calculation formula of the power at principle meridian is proposed. Two examples of progressive addition lenses with different additional power are designed, manufactured and tested under this formula. The power and astigmatism contours of the two lenses are presented. The results show that for the short-channel lenses with an additional power of 1.25 D, the near zone is reached 7-8 mm below the geometric centre and the maximum astigmatism in both blind zones is 1.25 D. For the longer-channel lenses with an additional power of 2.5 D, the near zone is reached 11 mm below the geometric centre, and the maximum astigmatism in the blind zones is 2.25 D. The principle meridian pattern can be applied to design the progressive addition lens with different addition power, same measurement point at reading area and various progressive corridor. It allows the wearers have more choices and comfortable visual effects, brings convenience for quality inspection in manufacturing process, cut down the production costs, thus improves the popularization of the of progressive lens.

The development of manufacturing has increasingly highlighted the importance of path planning in laser cutting technology. Traditional laser cutting path planning methods often face problems such as low efficiency and resource waste when dealing with complex and high demand tasks. Design a path planning method for laser cutting collaborative operation under immune algorithm. Selecting the feature points on the closed-loop contour of the parts as a set, transforming the path planning problem into a traveling salesman problem, constructing an objective function for single laser machine cutting path planning, and constructing an objective function for cutting collaborative operation path planning based on the single objective function. Develop path length constraints, material waste rate constraints, and total processing time constraints for the objective function of collaborative task path planning. Design a co evolutionary immune algorithm based on the concept of co evolution, and use this algorithm to solve the objective function of laser cutting collaborative operation path planning. The experimental test results show that the path obtained by the design method avoids intersections and does not pass through the cut area, thus obtaining the optimal path planning solution. Due to the implementation of multi laser collaborative operation path planning, the laser utilization rate has been greatly improved.

Due to the significant temporal variability, nonlinearity, and strong coupling of laser temperature characteristics, traditional control methods are difficult to meet precise control requirements. To address the above issues, a nonlinear self disturbance rejection control method for laser temperature is studied. This method utilizes the thermistor encapsulated inside the laser to collect real-time temperature of the laser and calculate the real-time error between it and the ideal temperature. Using error as input, a nonlinear self disturbance rejection controller is used to calculate the real-time control signal of the laser temperature, achieving precise control. The results show that under the control of the studied method, the coefficient of variation is relatively smaller, indicating that the method performs better in control effect and can better control the laser temperature to be closer to the ideal temperature, making the temperature more stable and reliable.

The precise positioning of the centroid of spaceborne laser spots can improve the performance and measurement accuracy of optical systems. In order to improve its positioning accuracy, a precise positioning method for the centroid of spaceborne laser spots under visual communication technology is proposed. Firstly, a CCD camera based on diffuse reflection in visual communication technology is used to capture real-time satellite laser spot images; Secondly, using the method of calculating the cross-correlation coefficient of light spots, the similarity evaluation of the collected light spot images is carried out to preliminarily determine the position and shape of the light spot; Finally, the improved Gaussian fitting sub-pixel positioning algorithm is chosen to refine the preliminary positioning results, thereby improving positioning accuracy and obtaining more accurate spot centroid positions. The experimental results show that the proposed method has a centroid positioning deviation parameter within 0.4 pixels in both the x-axis and y-axis directions, and the centroid positioning coordinates in multi frame spot images are relatively concentrated, indicating that the method has high accuracy and stability.

In the working environment of autonomous mobile robots, the distribution of obstacles has strong randomness, which puts higher requirements on the accuracy and adaptability of robot obstacle avoidance. To this end, an obstacle avoidance method for autonomous mobile robots based on embedded high-precision laser ranging is proposed. Firstly, using embedded high-precision laser ranging technology to collect environmental information and generate a grid map of the robot's moving environment. Then, the global optimal path of the robot is planned using the quadratic A*algorithm, and the optimal path is smoothed using the dynamic tangent method. Finally, during the process of the robot traveling along the global optimal path, the Morphin algorithm is used to plan local obstacle avoidance paths in real-time, achieving autonomous mobile robot obstacle avoidance. The experimental results show that in complex static obstacle environments, the planning of obstacle avoidance paths using this method takes less time, with an average time of only 1.129 seconds; In a dynamic obstacle environment, this method can successfully achieve smoother robot movement and obstacle avoidance. The time it takes for the robot to reach the endpoint is about 1 260 s, indicating that the obstacle avoidance effect of applying this method is more ideal.

As the core component of laser technology, the performance of lasers directly affects the effectiveness of applications. The beam quality is one of the important indicators for evaluating the performance of lasers. Studying the factors and their mechanisms that affect the beam quality of lasers is of great significance for improving the performance and application effects of lasers. Designing a mathematical model for the influence of relevant factors on the beam quality of lasers is crucial. Using M2 factor as the evaluation criterion for laser beam quality, ignoring the influence of coherent components, a Zernike spherical aberration coefficient impact model on laser beam quality is constructed. Construct a model of the impact of air thermal effect on laser beam quality based on the same correlation between the root mean square of phase distortion and the root mean square of light intensity distribution. Clarify the expression of the light field distribution of Gaussian beams at the orbit, construct a model of the effect of self focusing on the quality of laser beams based on the "Talanov" principle, and comprehensively construct a comprehensive mathematical model of the influence of relevant factors on the quality of laser beams by integrating three types of influence models. After adjusting each relevant factor one by one, the numerical simulation results of the model are completely consistent with the actual situation, proving the practicality of the model.

in recent years, with the development of new energy vehicle and lithium battery industry, the demand for welding copper materials is increasing, Welding of Non-ferrous metal materials such as copper alloys is already difficult to deal with because of the traditional infrared laser band and the peak absorption of copper materials, For Cu, Al and other materials, the absorption rate of blue light is 3-20 times that of infrared light, Deep penetration is achieved by using heat conduction mode in blu-ray laser welding, Compared with the infrared laser deep penetration welding mode, this paper adopts the single-tube blue laser beam combination technology with TO9 package structure. Sharp 5W 450nm high power blue light semiconductor single tube, The principle and optical path are designed by ZEMAX software, Solidworks is used to design the structure of high power blue light semi-conductor laser, Then through the Jun and precision machine six-dimensional real machine installation and debugging, 20 blue-ray semi-conductor laser with a power of 5 W and a width of 50 m based on TO9 packaging structure, Fiber output power of 86.4 W, to meet the basic sheet metal copper laser welding applications.

In the work of facial expression recognition in infrared thermal imaging, because infrared thermal imaging is easy to be disturbed by noise and the processing process is complicated, the time cost of facial expression recognition is relatively high. In view of this situation, an artificial intelligence-based facial expression recognition based on infrared thermal imaging is proposed. After collecting the infrared thermal imaging information, grayscale processing and denoising are carried out, and the two eyes are used as the benchmark to complete the normalization of the image. On this basis, artificial intelligence technology is adopted to segment the image according to the proportion of three chambers and five eyes, and facial expression features are extracted by using space-time descriptors, and facial expression features are described by histograms, and facial expression labels are made. Used to recognize facial expressions. The experimental results show that the proposed facial expression recognition method based on artificial intelligence takes less time in feature processing, the algorithm has high efficiency, and can maintain a high level of recognition rate under complex environment. It has good performance in practical application, and is suitable for application in practical projects.

The intelligent control of laser rust removal target trajectory for conventional electrical equipment mainly relies on laser vision detection technology, but this method lacks calculation of the curvature radius of laser beam data, resulting in a low fit between the final output trajectory and the expected trajectory, and the control effect is not ideal. To this end, an intelligent control method for the target trajectory of electrical equipment laser rust removal is proposed. Based on the principle of rust removal of power equipment using laser devices, the position coordinates of the beam centroid point are obtained to complete the calibration of the laser beam. The laser beam data is weighted and fused using the observation equation of the rust target, and a high-resolution sampling interpolation algorithm is used to calculate the curvature radius of the laser beam data. Based on this, a fuzzy controller with a dual closed-loop control structure is designed, Furthermore, intelligent control of the rust removal target trajectory can be achieved. The example application results show that the proposed method has a high degree of fit between the target trajectory and the expected trajectory, and the control effect is better. For the samples of power equipment with different corrosion degrees, the fitting degree of rust removal trajectory of the proposed method is higher than 98%, the average rust removal time is 10.1 min, and the energy consumption is between 1.7 and 1.8 kWh, which further demonstrates the application effectiveness of the proposed method.