Utilizing the photo-curing characteristics of photosensitive materials, a fiber-optic-like grating polymer long-waveguide structure was constructed between two multimode optical fibers through stepwise curing. This paper employed Norland Optical Adhesive (NOA) as the photopolymerization system. Initially, the optical adhesive was applied between two aligned fibers, followed by curing with a 405 nm laser to solidify the droplets, thus forming a polymer channel between the aligned fiber cores. By applying an external force to detach one end and repeating the operation, a grating-like structure was obtained. After forming the grating-like structure, a long waveguide was further cured to complete the grating-like polymer long-waveguide. The mechanical response of these waveguides was studied. By varying the dimensions of the gratings, the optimal grating polymer long-waveguide was identified. The results indicate that the introduction of the grating-like polymer long-waveguide leads to an extremely linear trend in optical transmission with waveguide spacing. Specifically, when the grating size of the grating-like polymer long-waveguide is 5 m, the linearity can reach above 0.99, with a strain sensitivity of 49/.

Laser cleaning technology has attracted widespread attention in both industrial and research fields due to its high efficiency and pollution-free characteristics. This review aims to explore the applications of different wavelength laser sources in cleaning technology, including carbon dioxide lasers, green lasers, ultraviolet lasers, as well as solid-state and fiber lasers. By delving into the characteristics of these laser sources and their applications in cleaning technology, the focus of future developments lies in multi-wavelength collaborative cleaning techniques and the advancement of intelligent cleaning devices, further enhancing the efficiency and sustainability of laser cleaning technology.

In order to meet the control requirement of multi-channel synchronous drive of frequency sweep light source in SS-OCT system, a frequency sweep light source control system based on FPGA as the main control unit is developed. The system uses Xilinx XC7A100T chip to design the main control module, five enhanced Howland current sources with buffer feedback to design the laser drive module, ADN8834 chip to control TEC to design the laser temperature control module, and host computer software to carry out human-computer interaction. Finally, the "wavelength-current" query table is made through the wavelength output automatic test experiment, and the automatic continuous linear output experiment is carried out by calling the "wavelength-current" query table. The system shows the control effect of 1nm control accuracy in the full wavelength range of the laser and 0.01 nm control accuracy in any wavelength range, which can be used for reference in the field of integrated control of frequency sweep light source of SS-OCT.

The existing intelligent metasurface electronic control systems fail to meet the rapid development requirements in terms of quantization accuracy, channel switching speed, adaptation for ultra-large arrays, and standard productization. This paper designs and implements a universal metasurface control system based on ZYNQ. The system consists of four sets of 128-channel high-precision programmable voltage output controllers cascaded together. A single controller utilizes the logic portion of ZYNQ to simultaneously control eight pieces of DAC, enhancing the switching speed. This setup enables independent voltage output for 128 channels, supporting continuous adjustable output voltage within the range of 0 to 20 V with an output accuracy 0.05 V and a response rate of less than 1.5 ms. By cascading the four sets of controllers, 512 channels of voltage output are achieved, facilitating intelligent metasurface dynamic control based on preset voltage encoding. Experimental results demonstrate that this control system meets the design specifications, offering high output accuracy, fast response speed, a high number of channels, and flexible, convenient usage, meeting the demands of intelligent metasurface control.

A high energy pulsed laser diode (LD) pumped Tm: YAG laser suitable for medical applications is reported. The oscillation operation and output characteristic of the Tm: YAG laser are studied by theoretical simulation and experiment. The results show that the pulsed Tm: YAG laser generates intense relaxation oscillation under pulse pumping state, and can output subpulses with high peak power. When pumped at an appropriate repetition rate, the superposition effect of reversed particles in multi-pulse pumping can significantly increase the single pulse energy. When operating with low-frequency pulses, the laser has a high threshold, which is not conducive to the output. In the experiment, the maximum single pulse energy is 476 mJ at a frequency of 120 Hz and a pulse width of 600 s, corresponding to an optical conversion efficiency of 26.4%. At an average pump power of 300 W, an average power output of 68.2 W is obtained, with an optical efficiency of 22.7%. The experiment of high energy pulsed laser output at 2 m above provides reference for laser medical applications.

In order to solve the problems of slow path planning speed, insufficient path safety and oscillation of motion speed in the autonomous navigation process of measurement robots in complex building decoration scenarios, a combined navigation path planning method that improves A * and optimizes TEB is proposed. To enhance the safety and effectiveness of the path planning for architectural measurement robots, collision evaluation detection for obstacles has been introduced. Additionally, redundant path points and unnecessary turning points in the global planning path have been optimized. The optimized TEB algorithm is employed for local planning between adjacent path points, achieving optimal path planning and improving the safety performance of the path while reducing planning time and minimizing path turning angles. The experimental results show that the improved A * algorithm reduces the path length by 23.51% on average and is 42.52% faster on average than the traditional A * algorithm in terms of planning time, and the addition of the optimized TEB algorithm achieves local path correction to achieve dynamic obstacle avoidance, which verifies the safety and effectiveness of the algorithm.

During the industrialized production of fabrics, fabric defects are varied and contain a large number of small target defects and elongated defects with extreme aspect ratios, which makes fabric defect detection a challenging task. To address this problem, an improved YOLOv5s algorithm is proposed in this paper. Firstly, the Mosaic data enhancement method is improved, which enriches the dataset while weakening the side effects of the original data enhancement method on the detection of some fabric defect types, and improves the detection of small targets and extreme aspect ratio defects. Then, the batch normalization is improved to representative batch normalization, which improves the algorithm’s differential feature representation of diverse defect instances and suppresses noise interference; finally, the lightweight coordinate attention is introduced, which encodes the long-distance dependency and channel dependency of the features with accurate location information, and enhances the algorithm’s ability to locate defects. The experimental results show that the algorithm in this paper significantly improves the detection ability of small targets and extreme aspect ratio defects, making the average detection accuracy mAP reach 81.3, which is 4.1% higher than the original YOLOv5s, and the detection speed is 32.6 fps, which fully meets the real-time requirements, and the algorithm better balances the detection accuracy and detection speed.

To quickly and accurately obtain the outer dimensions of small vehicles, improve the efficiency of motor vehicle inspection organizations, and reduce cost inputs, a high-precision measurement method for small vehicle outer dimensions based on RGB-D cameras is proposed. A hardware system for on-site measurements was built using a REALSENSE D455 camera and a Raspberry Pi development board. The target vehicles were extracted using the YOLOv8 model for object detection and instance segmentation. In view of the problems existing in RGB-D camera capturing vehicle images and generating point clouds, depth image repair and point cloud filtering are performed respectively, and the outline size parameters are automatically extracted after merging the point clouds. In experiments, outdoor field tests were conducted on four different models of small cars. The results indicate that the method’s measurement errors for vehicle outer dimensions is less than ±1%, meeting the requirements of national standards.

Fiber length measurement technology plays a crucial role in the field of light sensing, and how to accurately measure fiber length is a research hotspot in this field. The laser frequency modulation interferometry technique has the advantage of high spatial resolution measurement, which can be used in the field of fiber length measurement, but the frequency modulation nonlinearity will lead to spectral broadening, which affects the ranging accuracy. This paper proposes a fiber length measurement technique based on the Lomb-Scargle method, which can obtain more accurate measurement values while correcting the nonlinearity, and its measurement range is not limited by the optical range difference of the auxiliary path. As a verification, about 1.2 m long fiber length measurement experiments, the absolute error of the measurement of this method is basically stable within 50 m, with a high measurement accuracy. This method provides a way of thinking for the nonlinear correction of laser frequency modulation interferometry.

To achieve efficient change detection in remote sensing images, a lightweight Siamese network for change detection (LS-CDNet) was proposed. LS-CDNet is constructed based on a Siamese network architecture, with the lightweight network MobileNetV2 used as the backbone. A cascaded attention module is designed to optimize the low-level features and enhance the boundary information of the regions. In order to consider both the convergence efficiency of model training and the learning ability of imbalanced datasets for positive and negative samples, a weighted combination of the BCE Loss and Dice Loss is used to optimize the model learning strategy. Experimental results on the LEVIR-CD and CDD datasets demonstrate that LS-CDNet achieves precision rates of 88.06% and 90.12% respectively, with a model parameter size of 3.76 M and a computational cost of 2.18 G (FLOPs). The performance of LS-CD-Net outperforms other comparative methods.

Laser centerline extraction of line structured light is one of the key technologies in three-dimensional measurement of line structures light, and its extraction accuracy has an importantimpact on the measurement accuracy of the system. In actual 3D measurement, the image captured by the camera is affected by various factors such as the surface reflectivity of the object, which increases the difficulty of extracting laser centerlines. To address this issue, an improved extraction method combining grayscale centroid method is proposed. This method is based on connected domain analysis to remove background light bars and noise areas, and then implements local morphological processing based on linewidth distribution. Finally, the sub-pixel position of the laser centerline is calculated using the grayscale centroid method. The centerline calculation accuracy of the proposed method has improved by about 60% compared to traditional method. It can still accurately extract the position of the laser centerline in areas with overexposure and underexposure in stripe images, which can adapt to the test of complex environments and ensure the stability and accuracy of the laser centerline extraction work.

Aiming at the problem of low accuracy in autonomous driving systems with limited computing resources and multi-task driving perception algorithms, a multi-task driving perception algorithm with improved HybridNets is proposed. EfficientNetV2-S is selected as the backbone network of this algorithm to reduce the number of parameters, improve training speed and recognition accuracy; combine depth-separable convolution and use shuffle-channel convolution to reduce the amount of model calculation; use three independent decoders to solve problems of different difficulties, and add the A2-Nets dual attention machine block between the backbone network and the Neck end to fully extract global features. Compared with the basic network HybridNets, the mAP50 of this model can reach 79.8% in the vehicle detection task, an increase of 2.6%; the mIoU in the drivable area segmentation task can reach 91.8%, an increase of 1.2%; and the IoU in the lane line detection task can reach 32.55%, an increase of 0.93%. The running speed reaches 38 FPS. Experimental results show that compared with existing methods, the accuracy of the proposed method is greatly improved.

Aiming at the problems of imperfect point cloud feature extraction and low quality in traditional laser radar mapping algorithm, and the deviation of pose data caused by noise affecting the mapping effect, this paper proposes a laser mapping method based on improved Cartographer algorithm. Firstly, in the sensor information fusion part, the Adaptive Lossless Kalman Filter (AUKF) method was used to predict and update the sensor data, and then the noise was adaptively optimized to reduce the influence of noise on the pose data. Secondly, when processing the point cloud data collected by lidar, the effect of voxel filtering was improved, and the secondary screening of point cloud information was carried out by the method of point cloud weighting filtering to reduce the redundancy of point cloud and improve the quality of point cloud. Finally, the mapping test was carried out in the real environment, and the mapping effect of the improved algorithm and the traditional algorithm was compared. In the outdoor environment, the absolute translation error of the improved algorithm was reduced by 25.8% and the absolute rotation error was reduced by 28.9% compared with the original algorithm. It can be clearly seen that the data error of the improved algorithm is smaller and the mapping effect is more accurate.

The task of face image restoration can be achieved through the image-to-image translation. An improved face image restoration model is proposed based on the typical pix2pix framework in this paper. This model introduces perceptual loss and style loss on the pix2pix framework to enhance the generator's ability to handle image details and global consistency. Secondly, we integrate residual blocks in the network implementation process of the model to alleviate gradient explosion and increase the stability of the model. The experimental results show that the improved pix2pix model achieves better visual performance, with significant improvements in objective evaluation metrics such as PSNR and SSIM. These results demonstrate the effectiveness of the proposed model and provide a viable solution for the face image restoration task.

Due to the insufficient consideration of small and occluded target detection in existing methods, the tracking performance of targets is poor. Therefore, in response to the difficulty of point target tracking, a fast tracking method for point targets in sequence moving images based on 3D laser point clouds is proposed. Firstly, downsampling the 3D laser point cloud of a sequence of motion images and removing the ground data from it. By improving the Euclidean clustering method through dynamic thresholding, the image is segmented into targets and backgrounds. Then, the SECOND algorithm is improved, incorporating an adaptive spatial feature fusion module and a 2D convolutional neural network in the target detection stage. At the same time, 3D DIoU is used instead of Smooth L1 as the loss function to improve the target detection performance of the SECOND algorithm, Finally, the three-dimensional centers of each point target are constructed in the LiDAR coordinate system, and the 3D Kalman filter is used to continuously track the target. The intersection to union ratio and Euclidean distance are used as metrics, and greedy algorithms are used to match the nearest neighbor target to achieve fast tracking of point targets in sequence moving images. The experimental results show that the proposed method has a higher IoU value, reaching 0.971, and is more ideal for target tracking.

Aiming at the problem of target detection in hyperspectral remote sensing images, a machine learning based method for target detection in hyperspectral remote sensing images is proposed. Firstly, the dynamic evolution algorithm is used to find the projection direction that maximizes skewness and kurtosis, and high-dimensional image data is projected onto a low dimensional subspace to extract spectral information from the image. Then, the extracted information is transformed into histogram form through linear discriminant analysis, and the target area is initially segmented and classified using automatic labeling watershed algorithm and KNN method to remove non target spectral pixels. The test results show that after detecting and processing the target information, the overall classification accuracy and average classification accuracy of image pixels have a numerical interval of [0.97, 0.98], and the information entropy value is only 0.01, indicating that the method has high accuracy and high credibility of the results.

Laser remote sensing images are affected by environmental factors such as atmospheric conditions and the reflection characteristics of the target surface, resulting in blurring or loss of details in the images. To obtain high-precision single frame feature reconstruction results, a single frame feature reconstruction method in laser remote sensing images is studied. By calculating local correlation coefficients and fourth-order correlation coefficients in the low-frequency sub bands of laser remote sensing images, replacing the low-frequency coefficients of the original image with high-frequency coefficients, and performing non downsampling contour inverse transformation, the panchromatic sharpening of laser remote sensing images is achieved. The K-means clustering algorithm is used to group the image into bands, and independent reconstruction and iterative correction are performed. The reconstruction values of non reference bands are obtained through weighted average fusion. The experimental results show that the reconstruction accuracy of this method reaches above 0.95 and significantly improves the image detail effect.

There is complementarity between visible light images and infrared images in application, and fusion can improve the perceptual performance of images. Existing methods for fusion have poor image quality and cannot meet the application requirements of fusion images. Therefore, a visible light and infrared image fusion method based on generative adversarial networks is proposed. Build a visible light and infrared image fusion network framework using generative adversarial networks, and based on this, design a generator structure; Through brightness perception, adjustment, feature map extraction, and feature map fusion, visible light and infrared fusion images are generated. The discriminator network structure is determined, and the image fusion network is trained to make the distance between the true distribution and the false distribution approach zero until the constraint function - loss function - reaches the minimum value. The visible light image and infrared image to be fused are input into the trained network model, and the output result is the visible light and infrared fusion image. The experimental results show that the maximum mutual information value of the fusion image obtained by the proposed method is 17, the maximum average gradient value is 4, and the maximum edge strength value is 28.6, which fully confirms that the quality of the fusion image obtained by the proposed method is better.

Compared with ordinary space images, the laser three-dimensional barrier-free space image contains more information, which brings great difficulty to the image enhancement algorithm. In order to improve the quality of the laser three-dimensional barrier-free space image, the laser three-dimensional barrier-free space image enhancement algorithm under virtual reality technology is proposed. Under the virtual reality technology, the laser three-dimensional barrier-free space image is generated, and the combination of bilateral and guided filtering is used to denoise, and the image features are extracted. In addition, the enhancement processing results of the laser three-dimensional barrier-free space image are obtained through the steps of deblur, transformation enhancement, detail feature enhancement, color enhancement and so on. Through the test experiment, the conclusion is that compared with the traditional image enhancement algorithm, the optimized design algorithm shows that the signal-to-noise ratio and contrast of the enhanced image are higher, which increase by about 49 dB and 68 dB respectively, which proves that the optimized design algorithm has better image enhancement effect.

In the process of low resolution laser image processing, a sparse representation based method is used to achieve high-resolution image reconstruction. The obtained feature information is blurred, the peak signal-to-noise ratio of fuzzy features is high, resulting in poor image reconstruction effect. Therefore, a high-resolution reconstruction method for low resolution laser images under visual communication is proposed. Using sensitivity difference algorithm under visual communication technology to process low resolution laser images, the feature information contained in laser images has a high substantive effect. Using the quadtree partitioning idea to partition the preprocessed laser image, forming multiple image sub regions. Starting from each sub region image, input it into the Smooth PCANet for deep learning to extract deep level detail features of the image. Based on the sparse regularization model and combined with non local approximation prior constraints, high-resolution reconstruction of laser images is achieved. The experimental results show that using new research methods to process different types of laser images, the peak signal-to-noise ratio of the reconstructed images always exceeds 46 dB, and the structural similarity is above 0.96, achieving a significant improvement in image resolution.

The brightness and contrast of the image are usually low, which makes the target information blurred and increases the difficulty of recognition. Aiming at the problem that the traditional recognition methods are not accurate in the recognition of multiple salient objects, a new recognition method of salient objects in low-light images based on double-branch convolutional neural network is proposed. Image gray processing and denoising processing are implemented for low-light images. The low light image is enhanced, the low light problem is adjusted, and the salient object features of low light image are extracted by using double branch convolutional neural network. The experimental results show that: under the application of the proposed method, no matter how many significant objects exist in the image, the Kappa value is above 0.8, with high accuracy.

Low resolution laser image reconstruction has the problems of poor color visual effect and low structural similarity index. Therefore, a low resolution laser image reconstruction method based on color visual transmission is designed. Color visual transmission technology is introduced to fill the image color. Using ANC filter, amplitude is regarded as confidence, combined with bilateral filter and amplitude range kernel function, an adaptive bilateral normalization convolution method is designed to filter the image. Four-channel convolutional sparse coding is used to reconstruct low resolution laser images. The results show that the color visual transmission effect of the reconstructed images with this method is the best. The saturation is 97.2%, the brightness, hue, color contrast and sharpness are improved by 7.0%, 20°, 3.0 and 0.05 Line Pairs/MM, respectively. In addition, the visual area smoothness reaches 0.96, and the structural similarity index is 0.97. This method has better effect of laser image reconstruction.

The changes in lighting conditions in images can have an impact on edge detection. In cases of uneven or strong lighting changes, edge detection becomes more difficult. In order to achieve effective edge detection of optical images, a fuzzy mathematics based adaptive edge detection method for optical images is proposed. Using a compensation function to compensate for the illumination of optical images, enhancing their brightness, and then linearly adjusting and converting the brightness ratio of the images. By using homomorphic filtering methods to adjust the illumination and reflection components of optical images, the clarity of optical images can be improved. Combining wavelet transform method and adaptive dual threshold method to extract edge points from optical images, the fuzzy C-means method in fuzzy mathematics is used to cluster the extracted optical image edge points, obtain optical image edges, and achieve adaptive detection of optical image edges. The experimental results show that the DBI index of the proposed method is 0.21, the CHI index is 0.97, and the highest F-Measure value is 0.98, indicating good image processing performance and high edge detection accuracy, making it suitable for adaptive edge detection of optical images.

In order to effectively improve the channel transmission efficiency of optical communication systems, a research on channel modeling of free space optical communication systems in time-varying environments is proposed. This method is based on the analysis of atmospheric scattering and attenuation characteristics during signal propagation in optical communication systems in a time-varying communication environment. Using the atmospheric scattering coefficient and atmospheric attenuation coefficient as constraints, combined with the channel characteristics of the optical communication system, obtain the distance between the light source and receiver of the optical communication system, and calculate the signal power at the receiving end of the system. Establish a channel impulse response function based on the above constraints and parameters such as distance and power, and establish a channel transmission model for free space optical communication systems. The experimental results show that the method has a good fitting effect with the actual channel, and the channel model efficiency is as high as 0.9 or above. The channel model performance of this method is relatively high.

The deployment of fiber optic sensor IoT nodes is crucial for ensuring the efficient operation of the Internet of Things, but during the deployment process, issues such as low IoT coverage and high node energy consumption are prone to occur. Therefore, in order to effectively solve this problem, a deployment model for fiber optic sensor IoT nodes based on blockchain technology is proposed. Utilizing blockchain technology to construct a joint environment perception model for the Internet of Things, and obtaining perception information of fiber optic sensors in the Internet of Things environment; Construct an optimized deployment model for fiber optic sensor IoT nodes based on perceived environmental information, reducing node energy consumption and deployment time; The model is solved using the Non dominated Sorting Genetic Algorithm II (NSGA-II) algorithm to achieve the optimal deployment of fiber optic sensor IoT nodes. The experimental results of Retina Cortex show that the proposed method can improve the operational life of the Internet of Things, with a coverage rate of over 75% and a node residual rate of over 0.6%, indicating high deployment efficiency.

The divergence angle of laser is very small, and the beam quality directly affects the reliability and speed of communication, leading to redundancy of incremental big data. There is a problem of low residual load rate and high packet loss rate in laser communication network links. A method for selecting the optimal path for data transmission in laser communication networks based on big data analysis is proposed. Firstly, based on big data analysis technology, the beam quality increment of the laser communication network is calculated, and a routing decision function is established by combining the beam quality optimization cost, link transmission efficiency, link remaining load rate, and link bandwidth. Based on the decision function, the optimal path for laser communication network routing is obtained. Finally, performance testing experiments are conducted. The results indicate that the proposed method can describe the characteristics of routing changes in laser communication networks, with low link packet loss rate, high data transmission efficiency, and improved utilization of laser communication networks.

In order to ensure the security of communication data transmission in high efficiency spectral hybrid modulation MMwave fiber network communication, a method of secure data transmission in mmwave fiber network communication under double clustering algorithm is proposed. A hybrid modulation millimeter-wave optical fiber network communication channel model is constructed, and the dual functions of dimming control and efficient spectrum data transmission are combined to control the optical fiber network communication channel balance. An electromagnetic coupling modulation method is adopted to establish the multipath interference suppression model of the optical fiber network communication channel, the carrier modulation signal is introduced to adjust the output stability of the optical fiber network communication data, and the light intensity disturbance parameter is introduced to realize the efficient spectrum hybrid modulation and demodulation of millimeter-wave optical fiber network communication. The high-frequency feature quantity of optical fiber network communication security data is extracted, and the dual clustering algorithm is used to realize the block clustering processing of optical fiber network communication data, and the cascade channel path gain estimation of millimeter wave optical fiber network communication is realized. According to the parameter estimation result, the secure transmission of data is realized by combining channel modulation and anti-jamming design. The simulation results show that the bit error rate is 10-12, the signal-to-noise ratio is 49dB, the antiinterference ability is strong, and the transmission security and timeliness are good.

In order to reduce the bit error rate of weak light signal detection in laser communication, an underwater weak light signal detection method based on edge computing was proposed. The configuration management service in the edge gateway is used to set the configuration parameters of data collection service and edge computing service functions according to the configuration information issued by the cloud configuration server; Utilize data collection services within the gateway to collect communication signals emitted by underwater laser communication systems and upload them to the data server; The edge computing service in the gateway unloads the edge computing task during the acquisition and transmission of laser communication signals according to the multi-user computing unloading strategy of potential game, and speeds up the efficiency of laser communication signal acquisition and transmission; By calling the laser communication signals stored in the data server through the cloud platform, the bistable stochastic resonance method is used to detect weak laser communication signals. Experiments show that this method can effectively collect laser communication signals and complete the task unloading of edge computing; This method can effectively detect weak light signals in laser communication; At different signal-to-noise ratios and communication link lengths, the error rate of weak light signal detection using this method is relatively low.

In order to improve the utilization of laser communication channels and eliminate inter symbol interference, a channel equalization method for high hertz laser communication is designed. Based on the structure of laser communication network in the high Hertz frequency band, statistical mathematical methods are selected to construct a channel model of the laser communication network. Using the direct spread spectrum registration method in the high hertz frequency band, process multiple outputs of the channel equalizer and register the laser communication channel. The communication signal output by the channel equalizer is treated as a filter with finite pulse response, and a Toeplitz structure sensing matrix is constructed using pseudo noise sequences to estimate the phase of the laser communication channel. Based on the pseudo random sequence of the channel, the impact response of the laser communication channel is generated using the channel phase, and parallel communication of the channel is considered to achieve channel equalization processing in the high hertz laser communication network. The experimental results show that this method can achieve channel equalization in high hertz laser communication, with a communication error rate below 0.001 and superior channel equalization performance.

The demodulation precision of multi-channel grating sensing demodulation system is directly related to the degree of signal synchronization. In order to ensure the demodulation effect, the signal synchronization method of multi-channel grating sensing demodulation system is studied. Based on the application principle of multi-channel grating sensing demodulation system, it is determined that the transmission delay of adjacent gratings in the process of demodulation leads to overlapping of reflected optical signals and cannot guarantee the synchronization of multi-channel signals. After fully analyzing the signal delay of the demodulation system, Kalman filter method is used to complete the phase deviation compensation according to the frequency deviation compensation, and realize the multi-channel synchronous time compensation, so as to complete the signal synchronization of the multi-channel grating sensor demodulation system. The test results show that the time synchronization deviation is lower than 0.15 ms after the signal synchronization of multi-channel grating sensing demodulation system is carried out by the research method. The bit error rate of demodulation system after signal synchronization is lower than 0.55%. The maximum drift after demodulation is 0.007 8 nm. It can effectively avoid overlapping of reflected light signals.

In response to the problem that the elevation values between adjacent pixels in the digital elevation model are prone to discontinuity with the surrounding terrain features, which affects the data filtering effect, this study proposes an airborne LiDAR point cloud data filtering method based on elevation jump. The surrounding terrain features are an important basis for filtering airborne LiDAR point cloud data, and the elevation data between adjacent pixels is processed to achieve data filtering. This method first deeply analyzes the data characteristics of airborne LiDAR point clouds and the unique attributes of various ground objects, accurately assigns corresponding binary signals to each point cloud, and introduces them into the BP neural network to achieve fine classification of airborne LiDAR point clouds. Then, based on the results of point cloud classification, combined with the topological adjacency relationship of irregular triangulation networks, dynamic thresholds are set considering elevation factors and spatial angles to achieve filtering processing of ground feature data within any elevation range. Finally, virtual grid technology was introduced to process all point cloud data, combined with local adaptive threshold methods, to filter the remaining non ground point data. Experimental studies have shown that the proposed method has strong adaptability and can achieve good filtering effects on airborne LiDAR point cloud data, with greater application value.

In addition to narrowband interference signals, there are also other noise interference sources on the optical communication link. These noise signals may mask or interfere with the characteristics of narrowband interference signals, increasing the difficulty of identification. Therefore, a coherent optical communication link narrowband interference signal identification method based on wavelet transform is proposed. Obtain optical communication signals through optical communication link functions, and perform wavelet transform processing on the collected signals to remove signal noise while ensuring communication signal quality. Establish an expert knowledge graph of interference signals based on interference types, use convolutional neural networks to construct an interference signal identification model, input the communication signal processed by wavelet transform into the interference signal identification model for feature extraction, and combine the extracted features with the expert knowledge graph to determine the probability of the signal belonging to narrowband interference signals, thereby completing the accurate identification of narrowband interference signals in optical communication links. The experimental results show that using this method to identify narrowband interference signals in optical communication links has high identification accuracy and good identification performance, and has high practical application value.

In order to avoid problems such as trajectory deviation and obstacle collision during robot movement, a method for autonomous robot movement path tracking based on LiDAR is proposed. By using LiDAR technology to monitor the activity range of autonomous robots through radar, high-precision environmental contour information is obtained in real-time. Based on the obtained information, the SLAM map of robot movement is constructed. Use the PRM algorithm to plan the movement path in the environment, use a pure tracking algorithm to determine the control equation for the movement path tracking based on the planning results, and introduce an optimization equation for the tracking trajectory adjustment coefficient to achieve autonomous robot movement path tracking. The experimental results show that when using the above method to track the robot's moving path, the difference between the robot's moving speed, turning angle, and the target is small, indicating high tracking accuracy and good performance.

The effect of color space conversion is directly related to the quality of digital printing. In response to the problem of large color difference in traditional color space conversion, a color space conversion method for digital printing using laser imaging radar technology is proposed. This method is divided into two parts. In the first part, the laser imaging radar technology is used to obtain the printed manuscript image and remove scattered point cloud data from the image using neighborhood averaging. In the second part, the RGB color space of the printed manuscript is converted to Lab space, and then converted to CMYK color space using a neural network model. A simulation test was conducted on this method, and the results showed that after applying the method proposed in this paper, the image color difference was between 0.22 and 0.37, the structural similarity index was between 0.88 and 0.95, and the conversion time was between 21.9 and 25.6 seconds. This indicates that the conversion effect of this method is better, and it can better ensure the consistency between the color and the original printing color after conversion, while improving conversion efficiency and reducing color deviation.

With the rapid development of modern mechanical processing industry, its requirements for workpiece size accuracy are gradually increasing. However, there are still certain deficiencies in the dimensional accuracy of workpieces in the current mechanical processing process, and the methods for detecting workpiece size deviation are often affected by image noise and algorithm defects, leading to a decrease in the accuracy of workpiece size deviation detection. Therefore, a research on online detection method for size deviation of machined workpieces based on laser sensors is proposed. Select and configure laser sensors as the basis to obtain laser data for measuring the size of machined workpieces. Use the amplitude limiting filtering method to remove noise data from the laser data for measuring the size of workpieces, calculate the measured dimensions (length, width, and height) of machined workpieces, load the expected size data of workpieces, and measure the difference between the two. Based on this, determine whether there is a deviation in the size of machined workpieces, Thus achieving online detection of workpiece size deviation. The experimental data shows that the deviation detection results of the mechanical processing workpiece size (length, width, and height) obtained after the application of the proposed method are the same as the actual deviation, with a maximum error of no more than 0.5 mm and a distinguishable minimum deviation of 0.2 mm, fully confirming the effectiveness of the proposed method.

Laser triangulation can obtain the position of reflected laser light points based on the relationship between incident light and the normal of the measured surface. It is widely used in target recognition, displacement calculation, and other fields. Therefore, a face recognition method based on laser triangulation is studied to meet the requirements of accurate non-contact face recognition. This method uses laser triangulation to emit laser on the human face, obtaining different plane laser lines. Then, through the plane coordinate conversion method, the conversion relationship between the laser plane coordinate system and the image coordinate system is obtained. Then, a geometric structure model of facial recognition is established, and the geometric facial contour features of facial recognition are obtained using this model. Then, using the laser light point positioning algorithm, the facial recognition results are obtained by obtaining the local principal direction of the laser light point distribution. The experimental results show that the laser triangulation method is more accurate, and can effectively extract the contours of the mouth, nose, brow and eyes, and the accuracy of face recognition is higher. Even when the amount of data reaches 1 000 pieces, the recognition accuracy remains at 99.20%; When the amount of data reaches 1 000 pieces, compared with other comparison methods, the identification time of this method is the lowest, which is 0.21 s. In summary, the face recognition method based on laser triangulation provides an effective solution for security, intelligent interaction and other fields, and has broad application prospects and values.

The process of collecting target image data for intelligent manufacturing production lines is affected by noise, changes in lighting, and other issues, resulting in poor quality of input images, which in turn affects the accuracy of target detection. Therefore, a target detection method for intelligent manufacturing production lines based on 3D laser scanning technology is designed. Firstly, a 3D laser scanner is used to obtain the point cloud data of the target to be detected, and the topological relationship between the data is calculated using point cloud transformation criteria to generate a complete 3D laser image. Then, hybrid wavelet transform is used to denoise the 3D laser target image. Finally, four parameters, energy, entropy, contrast, and correlation, are used to extract image texture features and normalize them to create the optimal classification function. Support vector machine algorithm is used to partition the target sample image data of the production line, completing the intelligent manufacturing production line target detection work. The experimental results show that the proposed method has a maximum intersection to union ratio of 0.97 and an F1 value of 0.96, with an average detection time of only 0.53 seconds. This indicates that the proposed method has high detection accuracy, fast efficiency, strong robustness, and considerable usability in practical operations, providing technical support for the intelligent manufacturing industry.

Objective:To study the effect of YAG laser ablation on macular microcirculation. in physiologic vitreous opacity.Methods:The selected 32 patients of physiological vitreous floaters, all of them were treated by YAG laser ablation in our hospital,. Were checked for best corrected visual acuity (BCVA), non - contact intraocular pressure, ocular b-ultrasound, indirect ophthalmoscopy, to measure the foveal avascular zone and macular vascular density using optical coherence tomography angiography (OCTA) before and 30 minutes after surgery, 1 week after surgery, 3 months after surgery, and 6 months after surgery. Observe and compare the changes of the above checks.Results:BCVA and non-contact intraocular pressure had no significant difference between before and after treatment (P>0.05). The vitreous floaters obviously decreases, after treatment. No obvious morphological changes of macular arch ring were found, no microvascular abnormalities were found. There were no significant differences in blood vessel length density and perfusion density in the central, inner and complete areas of superficial macular area.Conclusion:YAG laser ablation can safely and effectively improve the symptoms of vitreous opacity, and has no adverse effects on the visual acuity, intraocular pressure and macular microcirculation

As one of the important components in the power grid, gas-insulated switchgear requires high manufacturing precision and gas tightness. At present, the gas-insulated switchgear filling box is mainly welded by manual argon arc welding and manual assembly, which is difficult to ensure the welding quality of the filling box due to the problems of large welding deformation and poor airtightness. In view of the above difficulties, the automatic tooling fixture and laser welding process for gas-insulated switchgear gas tank laser welding were developed. Through experimental verification, the developed automatic tooling fixture can ensure the welding quality and deformation control of the inflatable box; compared with the TIG process, laser welding can reduce the welding deformation of the inflatable box, and the weld quality is excellent, stable and reliable; using the experimental design method of process optimization, the optimal process parameters of the laser gas-tight sealing welding for laser power of 1 800 W, the welding speed of 20 cm/min, shielding gas N2, shielding gas flow 15-18 L/min, and finally, no leakage of gas in the inflatable box is found by helium mass spectrometry, and the sealing is good.

Traditional real-time target localization methods lack a certain depth in extracting feature information from target images, resulting in significant positioning errors. Therefore, a real-time target localization method combining laser point cloud data and image segmentation is proposed. Based on the characteristics of laser point cloud data, the data is processed through point cloud down sampling, radius filter denoising, and other methods. The image of the positioning target is symmetrically segmented, and features with similar attributes are clustered. The positioning target features are deeply extracted, and image matching is performed using time index based on the extracted positioning target features. The density based DBSCAN algorithm is used to cluster the coordinate matching data of the positioning target, Form a clustered 3D point set, convert the 3D point set to obtain real-time positioning coordinates, and perform real-time positioning of the target. The experimental results show that this method has a small deviation in the real-time positioning process, with an error of within 2.0%, and is effective.

In the digital delay signal processing of laser target shooting, it is often difficult for traditional filtering algorithms to effectively filter out noise interference, resulting in signal distortion. In order to solve this problem, this paper designs a digital delay signal filtering algorithm based on time-frequency peak. The signal-to-noise ratio of laser target digital delay signal is calculated, and the waveform equation of laser target digital delay signal is constructed to analyze the waveform characteristics of laser target digital delay signal. The instantaneous frequency is estimated by time-frequency peak, and the effective signal of time-frequency peak estimation is obtained. According to the standard variance of time-varying window, the confidence interval is divided, and the instantaneous frequency confidence range is determined to suppress noise interference. Two filtering criteria are set by energy and scale transformation, so as to complete the filtering of digital delay signal. The experimental results show that the range of signal amplitude fluctuation of the designed algorithm is [-10 dB~8 dB], which is consistent with the ideal range and has accurate filtering effect.

In the process of spot localization, if the quality of the laser triangular spot image is poor, it will directly affect the accuracy of subsequent spot localization. Therefore, a laser triangular spot localization method based on reinforcement learning is proposed. Collect laser triangular spot images and perform denoising and segmentation on the collected images to remove image noise, obtain spot targets, and improve image quality; Introducing neighborhood calibration light spots to perform cross correlation operations on the processed image, fully utilizing the skewed distribution of light spots and the similarity constraints of their positions, to determine the distribution of correlation coefficients at the target pixel level. Implement cubic non-uniform rational B-spline interpolation subdivision on the obtained correlation coefficient distribution results, establish the center positioning equation of the laser spot based on the subdivision results, optimize the positioning equation using reinforcement learning algorithm, further improve the positioning accuracy, and achieve precise positioning of the laser three corner spot. The experimental results show that using this method for laser triangulation spot positioning has good positioning effect and high accuracy.

In order to achieve accurate surveying and mapping of curved building plans and sections, 3D laser scanning technology is applied to the surveying and mapping work of curved building plans and sections, achieving optimization of surveying and mapping methods. Apply 3D laser scanning technology to obtain curved building point cloud data, and preprocess the initial point cloud data through steps such as fusion, denoising, and filtering. Extract the contour lines of curved buildings, and through two steps: measuring the structural data of curved buildings and drawing the horizontal and vertical sections of curved buildings, output the surveying results of the horizontal and vertical sections of curved buildings. Through the application effect testing experiment, it is concluded that the measurement error of geometric parameters of curved buildings is significantly reduced through the application of 3D laser scanning technology, and the similarity between the drawn horizontal and vertical profiles and the actual curved building structure is higher. Therefore, 3D laser scanning technology has good application value in the surveying and mapping of horizontal and vertical profiles of curved buildings.