A vortex beam is a light field with helical phase, which is widely applied in optical manipulation, optical sensing, optical communication, microscopy and quantum entanglement. The concept and generation methods of vortex beams are introduced. The recent advances of continuous vortex beams, pulsed vortex beams, nonlinear vortex beams and vortex microlasers in the past five years are reviewed, and their development trends are analyzed and prospected.

As the core of maritime transportation and an important military target in maritime warfare, the detection and recognition of ships have extremely profound practical significance in both civilian and military fields. However, the complex marine environment makes ship target detection difficult. Infrared imaging and target detection have excellent applicability in complex sea conditions. In response to the current demand for ship target detection, this paper summarizes the current development status of ship target infrared detection technology, analyzes infrared image preprocessing methods and ship target detection algorithms based on deep learning, compares the advantages and disadvantages of mainstream methods, and finally systematically summarizes and sorts out the research status and technological improvements, And prospects for its future development trends were presented.

Recently, visible light communication (VLC) technology has developed rapidly. This technology has the dual functions of low-power indoor lighting and high-speed broadband signal access, so it has high application value in the field of indoor big data communication. Multiple-input multiple-output (MIMO), as a technology that can achieve channel capacity multiplication in VLC systems, is increasingly attracting the attention of the industry. However, the strong correlation of multi-channels can easily lead to information crosstalk between users. The application of interference coding technology in transmitters can avoid information interference between users to a certain extent, and can improve the quality of system transmission and reception at a lower cost. This paper reviews the application background and basic principles of precoding technology in MIMO-VLC systems, summarizes the research progress of typical linear and nonlinear precoding technologies, and discusses and analyzes the application advantages, development trends and challenges of such technologies.

The near-infrared signal of the spaceborne solar spectrum monitor is usually weak and susceptible to external interference. To precisely detect the signal, a lock-in photodetection system based on InGaAs infrared detector was designed. The noise characteristics of detection system were analyzed, and it is found that suppressing the preamplifier noise can significantly reduce the overall noise. The PCB layout of preamplifier was optimized. Guard rings, a guard plane and a via fence were placed in order to strengthen the guard and shielding of input photocurrent signal. A tungsten lamp was used to simulate the solar spectral intensity in orbit and ground tests were conducted. The results show that the response linearity of the photodetection circuit to the incident light intensity is 99.95%. The signal-tonoise ratio (SNR) of each wavelength within the band is significantly improved compared with the previous solar spectrum monitor.

The chromatic confocal displacement measurement system, which has high accuracy and stability and can realize non-contact measurement, has been widely used in many fields such as industry, medical care and scientific research. At present, the factors restricting the application and development of chromatic confocal displacement measurement system are large range and high linearity. In this paper, the conditions of realizing linear axial dispersion of dispersive objective lens in a chromatic confocal displacement measurement system are discussed, and a kind of dispersive objective lens with high linear dispersion and large dispersion range is designed, and the appropriate tolerance range is specified, and a group of dispersive objective lenses are processed and assembled. A chromatic confocal displacement measurement system was established, and the displacement was calibrated by a dual-frequency laser interferometer, and the thickness of the standard gauge block was measured. The results show that the working band of the system is 450~650 nm, the range is 10 200 μm, the linear fitting judgment coefficient R2 reaches 0.993 7, and the maximum displacement error of actual measurement is only 3 μm, which proves that the chromatic confocal displacement measurement system has the characteristics of large range, high linearity and high measurement accuracy, and has a good application prospect.

In this paper, the spatial and temporal distribution characteristics of the crosstalk phenomenon with 532 nm continuous laser-irradiated surface-array CCD were investigated. Taking the three most upstream, midstream and downstream parts of the vertical transmission channel of the CCD as the typical irradiation sites. The results of the study show that laser irradiation of the CCD at different positions in the direction of the vertical channel does not affect the timing of the appearance of the main spot. As long as the moment of the beginning of the continuous laser irradiation or the moment of the end of the irradiation is before the readout transfer action, the main signal charge will form the main spot in the interference image. However, the irradiation site has an important influence on the spatial and temporal distribution of crosstalk lines. The crosstalk line length and spatially distributed position, at the end of the laser irradiation before the readout transfer moment, are related to the position of the vertical transport potential carrying the overflow charge arriving at the readout transfer moment. At the end of the laser irradiation after the readout transfer moment, it is related to the position of the bound pixel corresponding to the vertically transferred potential well that acquires the overflow charge. The results of this paper further refine the study of interference effects in continuous laser-opposed array CCD imaging devices.

This paper presents a single-layer transmission-type metasurface based on geometric phase, which enables the generation of broadband Terahertz vortex waves through the design and arrangement of metasurface unit structures. The metasurface unit consists of a top layer with dual C-shaped metal rings and a bottom dielectric layer. By utilizing the geometric phase principle and rotating the metal structures, the phase of cross-polarized transmitted light can be controlled while maintaining the same transmission coefficient over a wide frequency range. By encoding the unit structures with rotation, an encoded metasurface is formed capable of generating vortex beams with different topological charges in a wide frequency band. Simulation results demonstrate the generation of vortex beams with topological charges of l=±1 in the frequency range of 1.38-1.92 THz, with a maximum cross-polarization transmission efficiency of 23%. close to the theoretical limit of a single-layer structure.

A flexible surface-enhanced Raman scattering (SERS) substrate with a nano-bowl array structure was constructed and used for the analysis and detection of Rhodamine 6G and thiram. First, a single-layer polystyrene (PS) microsphere film was prepared on a glass slide using the gas-liquid self-assembly method. Then, the singlelayer silver nanoparticles obtained by the liquid-liquid interface self-assembly method were transferred to the PS microsphere film. Subsequently, polydimethylsiloxane (PDMS) was coated on the silver nanoparticles. After drying and peeling, an AgNPs@ PDMS nanobowl array substrate was obtained. Using a portable Raman spectrometer to test this substrate with Rhodamine 6G as the probe molecule, the detection limit was found to be 1×10-8 mol/L, with an enhancement factor of 4.32×105. The test results also showed that the substrate has high uniformity and reusability. Using thiram as the probe molecule, the detection limit was as low as 1×10-6 mol/L. The relationship between the Raman peak intensity of thiram and its concentration was studied, and the results showed that the substrate can achieve quantitative detection of thiram.

Laser remote sensing detection technology has the advantages of fast information transmission speed, high precision and long-distance detection, which can well make up for the shortcomings of seismic exploration in areas with complicated terrain, such as difficult construction and slow seismic data acquisition. In the previous study, the wavefront sensor, with its high sensitivity, high detection efficiency and off-axis detection, was used in the laser remote sensing detection of seismic waves, and a laser remote sensing detection system was constructed to explore the relationship between the ground amplitude and the spot measured by the wavefront sensor. In order to carry the detection system to the UAV and achieve the transformation from static detection on the ground to dynamic detection in the air, this paper designs an adaptive optical antenna based on FPGA, aiming at optimising the reception effect of the laser remote sensing detection system on the echo signals. The adaptive optical antenna achieves adaptive tracking and zoom by negative feedback adjustment mechanism, and makes real-time adjustment according to the received signal quality.

To improve the accuracy and resolution of small ptoelectronic encoders under the dual limitations of external dimensions and the number of code disc lines, a encoder subdivision method based on Coordinate Rotation Digital Computer (CORDIC) is proposed. Analyze the advantages and disadvantages of numerous electronic subdivision methods at present, analyze the causes of errors based on the subdivision principle, and use the improved CORDIC algorithm to perform high-precision subdivision processing on signals with less than one cycle of motion. The experimental results show that, compared with other methods, the maximum and minimum peak valley differences are reduced by 60″, 20″ and 10″ respectively, and the Root-mean-square deviation is reduced by 77.1%, 59.2% and 36.4% respectively, realizing the high-precision and miniaturization of displacement measurement.

In the process of collecting laser self-mixing interference signal, it is interfered by environment and circuit noise, resulting in signal distortion. In order to remove the noise and preserve the original signal features to the maximum extent, a self-mixing interference filtering method based on deep learning is proposed, which is suitable for weak feedback conditions. An autoencoder is used as a neural network, and a noisy signal is used as input and an unnoisy signal as output to train the network. The simulation results show that this method can not only improve the signal-to-noise ratio of the noisy self-mixing interference signal, but also preserve the waveform characteristics of the interference fringe, namely, the inclination direction of the fringe. In the experiment, the deep learning method is used to filter, and then the fringe counting method is used to reconstruct the displacement. The results show that this method has a good filtering effect on the self-mixing interference signal under the weak feedback condition.

Aiming at the problem of information loss in image feature representation of loop closure detection, a feature extraction algorithm based on Vision Transformer (ViT) with convolutional neural network for multi-model fusion was proposed. Firstly, feature extraction was carried out on the input image, and then the high-dimensional image feature vector was reduced by kernel principal component analysis (KPCA) to construct a new image feature representation. At the same time, a new range-matching algorithm was proposed, which limited and selected the range for feature matching through the corresponding range framework. The experimental results show that the proposed algorithm compared with other algorithms has higher accuracy and matching rate, and achieves better robustness and real-time requirements, which proves the effectiveness of the proposed algorithm in loop closure detection.

In the face of complex road traffic scenes, intelligent, fast and accurate detection of road identifiers is of great significance to automatic driving technology. The YOLOv7 algorithm with fast detection speed and high accuracy is suitable for real-time complex road identifier detection. In this paper, the YOLOv7 model is trained with the Chinese traffic sign detection dataset, and three different road traffic scene images, namely, ordinary, occluded and blurred, are selected to test the training model and compared and analyzed with three popular target detection algorithms, namely, CenterNet, Faster R-CNN and SSD. The results show that the YOLOv7 algorithm is fast in detection, has the highest average accuracy with 89.7% mAP, and has the best performance in the image tests of the three scenarios, successfully detecting road marker targets in the images even in the presence of occlusion and blurring.

With the continuous development of the economy, a large amount of waste aluminum alloy material has been generated in the industrial construction sector. The classification and recycling of waste aluminum alloy materials can enhance the utilization efficiency of waste resources and alleviate energy tension. This paper selects five types of aluminum alloys commonly used in the industrial field to investigate the influence of filament-induced breakdown spectroscopy (FIBS) spectral preprocessing wavelet transform basis functions on the classification accuracy of aluminum alloys. The orthogonal wavelet basis functions bior2.2, bior2.4, and bior2.6 are respectively used for preprocessing the FIBS spectrum of aluminum alloys, and the rapid classification identification of aluminum alloy types is achieved by combining with linear discriminant analysis (LDA), grid search optimized support vector machine (GSSVM) and back propagation neural network (BPNN). The results show that the average recognition accuracy rates of aluminum alloy types achieved by orthogonal wavelet basis functions bior2.2, bior2.4, and bior2.6 combining with LDA - GSSVM are 90%, 100%, and 76.67%, combining with LDA-BPNN are 96.67%, 100%, and 90%, respectively. Therefore, choosing appropriate orthogonal wavelet basis functions for FIBS spectral preprocessing methods and classification algorithm plays a significant role in improving the recognition accuracy of aluminum alloy types.

Due to uneven lighting in complex lighting environments, the intensity of light on the road surface varies, resulting in changes in the brightness and color of lane lines, making lane line detection difficult. In order to improve the accuracy of lane detection, a lane detection algorithm for complex lighting environments is proposed. The algorithm is mainly based on the texture features of lane lines, and uses voting algorithms to extract feature points from the vanishing points of images on traffic roads, constructing key detection areas for lane lines; Design an ant colony algorithm that combines gradient and statistical average relative difference for lane edge feature extraction, and finally fit and label the lane line to be tested. The experimental results show that the proposed method has high detection accuracy and strong robustness in complex lighting environments, significantly improving the success rate of lane detection under complex lighting conditions.

To meet the high precision requirements of structured light 3D measurement, a subpixel extraction method for laser stripes based on cubic spline interpolation and Gaussian fitting is proposed. Preprocessing including region of interest extraction and Gaussian filtering was first performed on the laser stripe images. The laser stripes were then thinned to obtain the initial center points using morphological methods. The normal direction of the laser stripes was obtained using the five-point fitting method. Fitting points were taken at subpixels in the normal direction of the laser stripes. The pixel values of subpixel coordinates were obtained by cubic spline interpolation. Finally, Gaussian fitting was performed to obtain the subpixel centers of the laser stripes. Experiments show that compared with the previous algorithm, this method has higher extraction accuracy and stability.

Aiming at the problems of missed detection, occlusion and low accuracy of target detection methods in foggy scenes, a foggy target detection algorithm YOLO-CL-CA based on multi-scale feature fusion is proposed. Firstly, in the data pre-processing stage, the AOD-Net model is used to defog the RTTS dataset to improve the image detail information. Secondly, a centralized feature pyramid CFPNet (Centralized Feature Pyramid) is introduced to regulate the shallow features with deep features to capture the key local regions of images and enhance the image feature utilization capability of the model. Thirdly, the CA attention mechanism (Coordinate Attention) is added before the output layer to improve the model’s ability to capture small target features. Finally, the LKC3 module is constructed by combining large convolution kernel to improve the problem of missed detection due to occlusion. The experimental results show that, the accuracy and mAP0.5 of the proposed algorithm are 90.6% and 81.7% respectively, 4.2% and 1% higher than these of YOLOv5s, which proves that the improved algorithm is effective and practical for fog target detection.

In the process of driverless obstacle detection for new energy vehicles, it is difficult to collect data due to various obstacles on the road and complex weather such as rain, snow and haze, resulting in low accuracy of driving obstacle detection. To this end, a method for detecting obstacles in unmanned driving of new energy vehicles using onboard LiDAR is proposed. Adopting a horizontal installation method, two 32 line LiDARs are symmetrically arranged on unmanned vehicles to collect 3D point cloud data of road structure and synchronize the data with coordinate conversion. A discrete point filtering algorithm based on statistical features calculates the standard deviation and global distance average of point cloud datasets to remove discrete noise points. Using Density Based Spatial Clustering of Applications with Noise (DBSCAN) clustering algorithm to achieve obstacle detection for autonomous driving of new energy vehicles. The experimental results show that the proposed method has high noise resistance, an average recognition distance of 82 m, obstacle detection accuracy of over 92%, and the highest false detection rate of 1.8%. The detection time is only 1.09 ms.

In order to solve the problems of under-utilisation of edge detail features in road extraction, and the difficulty of achieving accurate segmentation of roads in complex background occlusion regions, the study proposed a remote sensing road extraction model based on edge-guidance and multi-scale perception U-Net (EMUNet. Based on U-Net, the Canny edge detection result of remote sensing image is added as input, and the feature guidance of each layer encoder is carried out by designing the edge-guided fusion module combined with the attention, so as to make full use of the edge information to improve the quality of the final road extraction; secondly, in view of the background occlusion problem existing in the image, the multi-scale parallel hollow convolution module is constructed to enhance the multi-scale perception ability of the network, so as to capture more road information and to improve the quality of the road extraction. Secondly, to address the problem of background occlusion in the image, the network is enhanced by constructing a multi-scale parallel dilated convolution module to capture more contextual information and accurately extract the regions that are obscured by the background. The experimental verification is carried out on the Massachusetts road dataset, and compared with U-Net, EMUNet can achieve more accurate segmentation of small roads and occluded regions, and the recall rate, F1 score and intersection ratio are better than other comparative algorithms, so it can achieve more complete and accurate road information extraction.

Light field image can simultaneously record light intensity and direction information in the scene, but it requires more storage and transmission resources, so that efficient encoding schemes are required. This paper proposes a light field image coding method based on human visual system, achieving end-to-end perceptual light field coding. Considering that under the same distortion level of light field encoding, the distortion in salient region is more easily perceived, a light field image perception coding method is proposed. Aiming at the perception redundancy of sub aperture images, considering that the Just Noticeable Distortion (JND) model in pixel domain does not pay attention to the salient region characteristics, a JND masking model combining saliency is constructed; and then, adaptive rate allocation considering human perception is performed for each sub aperture image frame. The experimental results show that in the low latency mode, compared with the High Efficiency Video Coding (HEVC) encoder, the proposed method saved an average bit rate of 16.01% under the same quality index of Significance Weighted Peak Signal-to-Noise Ratio (SW-PSNR), it improves the efficiency of light field image encoding while achieving a higher degree of fit with the subjective perception of the human eye.

In recent years, light field images have received great attention as a new form of digital images. Light field images have the characteristics of high dimensionality and large amount of information, and the pressure on storage and transmission bandwidth has become a bottleneck in their development. In this paper, method for constructing and compressing optical field based on adaptive matching pseudo-sequence is proposed. Firstly, the original image is resampled based on the center point of the microlens to obtain the aperture image of the optical field; Then, by utilizing the correlation between adjacent views of the light field image, a pseudo-sequence is reconstructed to adaptively determine the optimal encoding order of the light field image that conforms to the multi-functional video encoder for different light fields; Finally, perform video compression. Experimental results show that this algorithm has better compression performance than VTM platform, BD rate reduces by 13.62% on average, which is better than the other comparison algorithm.

Aiming at the problem of confusion between foreground class and background class caused by the lack of example labeling in few-shot remote sensing images, a object detection method for few-shot remote sensing images combined with pixel attention and decoupled classifier is proposed. In this method, a novel pixel attention feature pyramid structure is designed to capture more important spatial and channel semantic information, so as to highlight the key features of the objects while suppressing background noise. In addition, the standard classifier is decoupled into two parallel detection heads to process the foreground classes and the noisy background classes respectively to alleviate the bias classification problem of the classifier. The proposed method is experimentally carried out on two public remote sensing datasets, and the results show that compared with the current new methods, the average accuracy of the proposed method is improved by 4%-7% on the DIOR dataset and 11%-17% on the NWPU VHR-10 dataset, and the detection performance is good.

In order to fully consider the spatial spectral structure features of hyperspectral remote sensing images, reduce data redundancy, obtain more recognizable features and improve classification accuracy. In this paper, we propose a fractional-order Gabor-based hyperspectral image classification method, which implements multi-resolution analysis of local signals in the fractional domain to enhance the characterisation of hyperspectral images. Firstly, a multiple component fractional-order Gabor filter is constructed by setting up a multiple order sine wave to obtain an effective feature representation. Secondly, the Gabor phase features are encoded by quadrant bits, and the code distance is calculated by Hamming distance, which reduces the computational complexity. Finally, the Gabor phase features of different orders are fused to obtain complementary texture information in order to obtain higher classification performance. Based on the Trento real dataset, three classification samples were selected for training. The overall classification accuracy reached 87.15%, and the Kappa coefficient was 0.83. The experimental results have verified the effectiveness of this method in small sample training, and compared with other algorithms, it has improved classification accuracy.

There is a problem of laser scattering in the image acquisition process of laser images, which affects the image effect. Therefore, a method for repairing laser images with backscattered light under visual communication technology is proposed. The collector collects the laser image, denoises the laser image through the sparse denoising method, applies the adaptive color correction factor in the gradient domain Tone mapping method to correct the color of the laser image, introduces the linear color correction model to eliminate the color difference of the image, and completes the back scattered light repair of the laser image based on the improved Generative adversarial network. The experimental results show that the repaired laser image using the proposed method is consistent with the actual situation of the experimental object. The similarity of the feature structure between the repaired image and the original laser image is more than 90%, and the repair time is less than 89.5 ms, indicating good repair results.

Laser spot images are easily limited by imaging conditions and imaging methods during the imaging process, resulting in low resolution of laser spot images and difficulty in meeting practical needs. Therefore, a super-resolution reconstruction method for laser spot images based on visual communication technology is proposed. The visual communication technology is used to collect the laser spot image, and the dual tree complex wavelet threshold method is used to denoise the laser spot image. The dense neural network is improved to extract the characteristics of the laser spot image, the number of atoms in the dictionary is reduced based on the singular value decomposition method, and the sparse expression regularization method is improved to achieve the super-resolution reconstruction of the laser spot image. The experimental results show that the low resolution image reconstruction results of the proposed method are closer to the original image, and the structural similarity of the reconstructed images is above 0.9, proving that the reconstruction effect of this method is good and more suitable for practical applications.

3D object point cloud recognition is an important component of environment perception tasks for intelligent robots. Based on dynamic graph features, this paper proposes a Dynamic Graph Stacked Broad Learning System (DG-S-BLS) network for 3D object recognition. DG-S-BLS extracts high-dimensional features from point clouds using a dynamic graph convolutional network, and then uses the Broad Learning System (BLS) model to classify point clouds based on the overall features of samples. The classification accuracy is further improved by using the Stacked BLS model performed upon the residual of the BLS blocks. Experimental results on the LiDARNet outdoor point cloud dataset show that the classification accuracy of DG-S-BLS reaches 99.5%.

In order to solve the UAV cooperative operation problem, multi-tasks need to be assigned to multiple UAVs. Wireless ultraviolet light is used to realize the covert information transmission between UAVs under the strong electromagnetic interference environment, and an improved particle swarm algorithm for multi-UAV task allocation is proposed, which takes into account the threat cost, voyage cost, and the time difference of completing the task for UAVs to perform the task, and combines the compression factor and the differential evolution idea to solve the problem that particle swarm optimization algorithm is easy to fall into the local optimum. Simulation results show that the improved particle swarm algorithm improves the average success rate of task allocation by about 16% compared with the traditional particle swarm algorithm under different ratios of UAVs and number of tasks, reduces the number of iterations of the algorithm by about 4.5 times on average at the time of convergence, and the optimal fitness value decreases by nearly double on average, which is of some significance in the practical application of multi-UAV task allocation.

The wavefront correction of adaptive optics is vulnerable to the interference of complex background, light intensity, noise signal and other problems, resulting in reduced correction effect. In order to solve these problems, a adaptive optics wavefront correction technique based on deep learning for airborne remote sensing communication system is proposed. The adaptive optics wavefront is detected by the phase difference method, and the noise is eliminated by the wavelet transform algorithm to avoid the interference of noise in the correction process. According to the prediction and self-learning ability of the deep neural network, a dynamic model network, a strategy network and a decision unit are constructed. By comparing with the correction threshold, the adaptive optics wavefront of the airborne remote sensing communication system is corrected. The experimental results show that the proposed method has a Stellerian ratio close to 1, and has a short correction time and good correction effect.

A distributed fiber grating signal demodulation technology based on Labview is proposed due to poor channel balance caused by inter symbol interference in communication channels in distributed fiber grating signal demodulation. The continuous time sampling technology is used to build a distributed fiber grating signal acquisition model, and the power spectral density characteristics of the signal output are phase modulated. The offset center frequency and single sideband power spectral density characteristics of the distributed fiber grating signal are extracted under the discrete Wiener model, and the signal is modulated and demodulated through the modulation and demodulation technology of Labview, Finally, the constellation analysis method of resisting phase noise interference is used to analyze the anti-interference and error suppression capabilities of fiber grating signal demodulation. The simulation results show that this method has good anti-interference performance for distributed fiber Bragg grating signal demodulation, improves the spectral bandwidth of the channel, reduces frequency center offset, has a low transmission error rate of 0.05%, and maintains a signal transmission delay of 30 μs, with strong applicability.

Due to the increasing bandwidth requirements and network size and complexity of optical communication transmission, data transmission paths have more selectivity, which can easily lead to high energy consumption and low efficiency of routing. Cloud computing platforms can provide massive computing resources and storage space, and can handle large-scale datasets. Therefore, based on cloud computing intelligent embedded technology, a type of optical communication network routing is studied. Design the signal conversion, encoding, decoding and other functions of optical communication networks using embedded technology, and establish a network routing energy consumption model based on the energy consumption characteristics of data transmission and network nodes; Introduce ant colony algorithm and find all routes that can achieve data transmission based on the principle of ant foraging; Using genetic algorithms, the best route for data transmission among these routes is identified through selection, crossover, and mutation calculations, that is, the route with the least energy loss. The experimental results show that the proposed technology can improve data transmission speed, achieve a packet loss rate below 3.0%, and a communication delay of only 23 ms, reducing the energy consumption of data transmission network nodes, confirming the good routing communication performance of the proposed method.

The distribution of nodes in ultra large capacity fiber optic communication networks is relatively chaotic, and the correlation between nodes is complex, resulting in low accuracy in security domain division, seriously threatening the security of fiber optic communication networks. To this end, a security domain partitioning method for ultra large capacity fiber optic communication networks is proposed. According to the fiber optic network environment, the K-means algorithm is used to set the initial security center value, calculate the distance between the representative sample value in each region and this value, and solve for the average standard values of the maximum distance and the shortest distance. On this basis, the particle swarm separation method is used to identify intrusion signals, and the changes in membership values between each node and the set local area are calculated separately under normal and presence of intrusion particle interference, in order to determine the phase difference of the signal. Based on the correlation value between nodes, the node area that matches the dissimilarity inverse vector is divided to complete the security domain division. The experimental results show that the proposed method has high accuracy in partitioning, and after the implementation of the algorithm, the number of nodes participating in attack events significantly decreases, and network security is strengthened.

More and more ONUs are connected to passive optical networks, which are prone to network congestion, information transmission delay, data packet loss, and other phenomena. Therefore, an orthogonal frequency division multiplexing passive optical network system based on information entropy characteristics is designed. The system framework structure is composed of optical line terminals (OLT), optical distribution networks (ODN), and optical network units (ONU). The system network structure is designed in a decentralized light splitting form, and the system database and data tables are designed using MySQL. Design three main functional modules of the system, namely the passive optical information transmission and reception module, the ONU load judgment module based on information entropy, and the passive optical network bandwidth allocation module. By reasonably allocating bandwidth resources, improve the transmission capacity of the passive optical network. The results show that the transmission time slot length of the passive optical network system is below 4 ms after 5 rounds of polling, and the communication transmission efficiency of the passive optical network is high, which can quickly achieve information transmission and greatly avoid delay problems.

Because the received signal is not de noised, the problem of large estimation error and low accuracy of delay estimation in optical fiber communication link is caused. In order to solve the problem in the method, a study on delay estimation in high-speed optical fiber communication link based on edge computing is proposed. The optical communication network architecture is designed through edge computing. The network communication model is built based on the cloud computing layer in the optical communication network architecture, and the optical signal is obtained. The optical signal is transmitted through the direct link, and the high-speed optical fiber communication model is established. The generalized correlation algorithm is used to denoise the pre filter of the received signal. After the received signal is denoised, the Kalman filtering method and Round trip algorithm are combined to complete the time delay estimation of high-speed optical fiber communication system. The experimental results show that the proposed method has high accuracy, good effect and high practical application value in high-speed optical fiber communication.

The weak light signal of underwater laser communication system is interfered by spectrum and multipath, and the detection accuracy is not high. In order to improve the stability of underwater laser communication, a detection method of weak light signal of underwater laser communication system based on deep learning is proposed. The spatial beam forming method is used to construct the underwater laser communication acquisition and multi-layer spectrum decomposition model, and the beam characteristics of the signal are enhanced. The radio frequency tag distribution information and multipath information of the weak optical signal are estimated by combining the time reversal technology, and the weak optical signal of the underwater laser communication system is enhanced by using multipath fading suppression. The optimization control is carried out in the detection process of the weak optical signal by combining the deep learning algorithm, so that the whole signal detection process is guaranteed to have a constant false alarm probability. The simulation results show that this method has a good ability to suppress multipath fading when detecting weak light signals in underwater laser communication systems. The error rate is low in different signal-to-noise ratio environments. When the signal-to-noise ratio increases to 90 dB, the error rate is only 10-5, and the detection time is less than 10 ms, which is superior to the comparison method and has greater application value.

In order to improve the wear and corrosion resistance of SUS630 stainless steel (SS) used as valve stem, Stellite 6 cobalt-based alloy coating was prepared on its surface by laser cladding technology. The microhardness, wear and corrosion properties of Stellite 6 coating and SUS630 SS were studied. The results show that compared with SUS630 SS, the hardness of Stellite 6 coating increases by 25% and the wear volume decreases by 50%, indicating that Stellite 6 coating has better wear resistance than SUS630 SS. In addition, in 3.5% NaCl solution, the corrosion potential (-0.18 V) and the impedance value of the passive film (367 kΩcm2) of the Stellite 6 coating were higher than that of SUS630 SS (-0.35 V and 129 kΩcm2), while the passivation current density (10-7 -10-6 A/cm2) of Stellite 6 coating was lower than that (10-6 -10-5 A/cm2) of SUS630 SS, indicating that Stellite 6 coating has better corrosion resistanc than SUS630 SS. Therefore, laser cladding Stellite 6 coating can improve the wear and corrosion resistance of SUS630 SS.

In order to study the effect of plate thickness on residual stress of GH625 superalloy subjected to laser shock peening, on the one hand, laser shock strengthening technology was used to impact 2 mm GH625 plate, and the surface residual stress distribution was obtained by X-ray stress tester; on the other hand, ABAQUS finite element simulation software was used to perform numerical simulation of laser shock GH625 plates with thicknesses of 2 mm, 3 mm, 4 mm, and 5 mm. Through the combined method of experiment and numerical simulation, the Johnson-Cook (J-C) constitutive equation of GH625 high-temperature alloy was optimized. The results indicate that when the strain rate influence factor C is 0.005 21, the simulated residual stress on the surface of laser shocked GH625 is in good agreement with the experimental results. As the thickness of the GH625 plate increases, both the magnitude of compressive residual stress on the surface and the depth of residual stress distribution increase. There is also compressive residual stress on the backside of the plate, which decreases with the increase of plate thickness.

There are many obstacles in the working environment of logistics sorting robots, so a research method for autonomous navigation of logistics sorting robots based on laser SLAM is proposed. Using laser SLAM to determine the distribution of obstacles in the robot’s working environment; Add the target bias strategy to the RRT * algorithm, and by setting fixed and variable probability target bias strategies, make the path nodes more directional when traversing the target points, better avoiding collisions between robots and obstacles, and then use the search random tree to search for the shortest path among them. Comparative experiments have shown that compared with conventional methods, this method can guide the robot to follow the shortest path between the starting point and the target point, and the navigation process takes less time, resulting in an error rate and energy consumption of about 5% for robot autonomous navigation.

A new method for measuring the center position of the thermal physics laser spot has been designed, as traditional methods can cause diffraction interference and result in deviation in the measurement results. Preprocess the thermophysical laser spot image, calculate the second-order differential in the horizontal numerical direction, calculate the threshold through the Gaussian function, realize image enhancement, extract the spot position features, obtain the position coordinates of the spot center, obtain the position coordinates of the thermophysical laser spot center based on the local gray centroid method, and realize the measurement of the thermophysical laser spot center position. The performance test results of the method indicate that the average standard deviation of the azimuth deformation angle measurement results and the elevation angle measurement results obtained by the designed thermophysical laser spot center position measurement method under various indexing table calibration values is 0.39 and 0.17, both lower than the three traditional methods, verifying the effectiveness of the method.

The detection accuracy is not high due to the interference of binocular vision coupling in high-resolution 3D lidar imaging target detection. A method of high-resolution 3D lidar imaging target detection based on deep learning is proposed. The vision sensor and inertial sensor are used to collect high-resolution 3D lidar images, and the method of laser point cloud feature extraction and dynamic scene image fusion is used to synthesize high-resolution 3D lidar. The compensation filtering method is used to correct the point cloud distortion in the process of 3D lidar imaging target detection. The depth feature matching model of lidar imaging target detection is established through the map feature matching and depth learning model of local sliding window, and the high-resolution 3D lidar imaging target detection is realized according to the feature matching results and the lidar frame matching results. The simulation test results show that the laser radar imaging using this method for target detection has a minimum peak signal-to-noise ratio of 41 dB under noise interference, and the detection accuracy reaches 0.96.

Through the design of autonomous lane changing lateral obstacle avoidance for intelligent connected vehicles, the safety performance of vehicle autonomous driving is improved. A laser point cloud based intelligent connected vehicle autonomous lane changing lateral obstacle avoidance method is proposed, and a vehicle driving environment excitation point cloud data collection model based on an all laser radar unmanned driving system is constructed. The perception of the intelligent network vehicle driving environment is achieved through the laser radar laser point cloud image detection method. Combining the methods of obstacle detection, lane line detection, and driving area detection, this paper analyzes the correlation characteristics of obstacles and lanes for vehicle lane changing under intelligent network, searches for boundary point through the autonomous optimization method of laser point cloud until the search path reaches the threshold, and then realizes the planning and design of lateral obstacle avoidance for vehicle autonomous lane changing based on the straight line road model and the continuity principle of lane lines. The simulation results show that the proposed method achieves convergence in 5s, the obstacle coordinate identification error distance is less than 0.7, and the identification accuracy of the proposed method is more than 0.970, which improves the perception ability of lane information, accurately and efficiently estimates the driving area, and improves the obstacle avoidance ability of vehicles.

Due to the complexity of the optical path of the laser 3D projection system and the influence of the shape and size of spatial targets, it is difficult to achieve ideal projection effects in a specific space, resulting in a significant deviation between the projection target and the actual projection point. Therefore, a radiometric illumination calibration method is proposed. Using two different shapes of radiation surfaces and sensitive surfaces, rectangular and circular, respectively, to construct radiation irradiance measurement instruments and obtain four different radiation irradiance measurement results; Using the Stefan Boltzmann total radiation law, calibrate the measurement results, and then complete the calibration of the irradiance of intelligent laser 3D projection space targets. The experimental results show that the radiometric calibration results of the proposed method are basically consistent with the actual results, and the radiometric calibration error remains within 6 W·m2.

The optoelectronic reservoir computing system based on delayed-feedback structure is a kind of analog optical signal processing system with simple-software and simple-hardware design. If it is successfully applied to the real-time recognition of ventricular premature beats, it is of great significance for the effective prevention and early treatment of ECG diseases. In this paper, an optoelectronic reservoir computing model was established using a Mach-Zundel modulator with an optoelectronic feedback loop, and the ventricular premature beat electrocardiogram was recognized. The key parameters of the model were analyzed and optimized. For example, the effects of feedback strength β, input gain γ, number of virtual nodes N on electrocardiogram recognition were analyzed. The best recognition normalized mean square error (NMSE) was 0.080 9, the best recognition rate was 94%, and the sensitivity was 94.9%. The results indicate that the optoelectronic reservoir computing system exhibits a good performance in ventricular premature beats recognition, and simplifies the complexity of current electrocardiogram recognition algorithms.

There are various types of surface defects in new energy vehicle components, and relying on manual inspection has problems such as false or missed inspections. In order to improve the accuracy of surface defects in new energy vehicle components, a surface defect detection method for new energy vehicle components based on machine vision technology is proposed. Combining Ridegelet transform and wavelet transform, denoising is performed on surface defect images of new energy vehicle components collected based on machine vision technology without damaging image details. MSR algorithm is introduced to realize fast image enhancement through adaptive calculation of information entropy proportion weight. The mathematical morphology calculation method is used to extract the edge of the main body of defects in the image of new energy vehicle parts, and then the surface defects of new energy vehicle parts are detected. The experimental results show that the proposed method has a detection time of less than 10 ms and accurate defect location analysis. In the face of surface defects in multiple types of new energy vehicle components, it can achieve accurate identification and improve the defect detection efficiency of new energy vehicle components.

CNC laser cutting machine is a large-scale equipment widely used in the manufacturing industry today. Due to the existence of vibration and deviation phenomena in the internal feed system, feed error cannot be avoided. If it is not compensated, it will affect the application effect of the cutting machine. Therefore, a high-performance CNC laser cutting machine feed error compensation method research is proposed. Thoroughly analyze the sources of feed errors in the cutting machine, determine the composition formula of feed errors (positioning error+force deformation error), clarify the measurement method of positioning error and force deformation error, obtain feed error data from multiple measurement points at the same time, use weighted average to calculate the final feed error to be compensated, and develop a feed error compensation program combining hardware and software equipment, By executing the established program, effective compensation for feed errors can be achieved. The experimental data shows that the feed error compensation results obtained by the proposed method are consistent with the expected feed position, and the minimum response time for feed error compensation is 0.1 seconds, fully confirming that the proposed method has stronger feed error compensation performance.