As an advanced surface modification technology ,laser cladding technology has the advantages of small coating dilution rate ,good combination of cladding layer and substrate ,and fast cooling speed. At the same time of the rapid heating and rapid cooling process ,the coating and the substrate often produce high residual stress ,resulting in adverse effects such as deformation ,coating peeling and cracks. This paper first analyzes the generation mechanism of laser cladding residual stress ,and then summarizes the numerical simulation research of laser cladding residual stress ,including model establishment ,distribution of single-pass and multi-pass cladding residual stress ,and the influence of process parameters on residual stress ,and the result verification ,and introduced the residual stress control method ,finally summarized the deficiencies and development directions in the numerical simulation research of laser cladding residual stress.

Cancer is the second leading cause of death worldwide and poses a significant threat to individuals ,families and society. Rapid and accurate diagnosis of cancer at an early stage is the primary means of mitigating the le- thal effects of cancer. With the development of biosensors ,optical biosensors have received widespread attention. This kind of sensor has the advantages of fast ,real-time ,accurate ,high sensitivity and multiple detection ,and has great application potential in the clinical diagnosis and treatment of cancer. This article discusses the advantages and disad- vantages of existing biosensors for cancer marker detection ,starting with cancer markers. The working principle of pho- tonic crystal fiber and surface plasmon resonance technology are introduced ,and the latest research progress of photon- ic crystal fiber optic fiber sensor based on surface plasmon resonance in early cancer diagnosis is reviewed ,combined with the literature reports of PCF-SPR sensors in the field of biology in recent years. Finally ,the progress of PCF - SPR biosensors in the field of cancer detection is summarized and prospected.

Acousto-Optic Tunable Filter ( AOTF) is a dispersive element based on acousto-optic effect. By chan- ging the ultrasonic frequency ,the diffraction direction of the light can be changed ,so as to select the wavelength. In this study ,a new two-pass dispersive AOTF structure is designed and studied using finite element analysis software. The results of the present study show that the design of this dual-range dispersive AOTF can effectively compress the spectral linewidth. When the center wavelength is 1 064 nm ,the linewidth compression efficiency reaches 28. 6% at an ultrasonic operating frequency of 84 MHz ,which indicates that the spectral resolution has been significantly im- proved ; at the same time ,this structure also has a strong inhibition of sidelobes ,which can compress the sidelobes of the diffracted light with a center wavelength of 1 064 nm from -27. 0 dB to -44. 8 dB ,which improves the spectral pu- rity.

A folded cavity mode-locked system has been employed by using the core device SESAM ( semiconduc- tor saturable absorption mirror) for related experiments. The 1 064 nm picosecond ultrafast pulse has been expected to be obtained in experiments. Its characteristics based on waveform and power have been analyzed. At the same time ,the influence of pump power and cavity length have also been explored. Finally ,the mode-locked pulse waveform with repetition rate of 92. 25 MHz has been obtained. It is shown that with a certain cavity length ,the mode-locked pulse will not start if the pump power is insufficient. If the pump power is too high ,Q-switched instability and multi-pulse phenomena will be observed. Within the range of pump power that leads to mode-locked effect ,the average power of the mode-locked pulse is positively corrected with it. When the pump power is constant ,the mode-locked pulse wave- form is very sensitive to the change of the length of the folded cavity. The change of the cavity length of less than 1mm will make the waveform jitter greatly ,and the stable comb wave will appear only in a very small cavity length range.

Traditional Gray-code assisted four-step phase shifting phase unwrapping method requires projection of four stripe patterns ,but the reconstruction accuracy is not high due to factors such as environmental noise and defocus- ing of the projector. In order to improve the efficiency and accuracy of three-dimensional reconstruction ,a phase un- wrapping method combining Hilbert transform and Gray code is proposed. First ,the cosine component is obtained u- sing the acquired sine stripes image and the background light intensity image ,and then the cosine component is con- volved with the shock response using the Hilbert transform. Shifting the phase by π/2 while keeping the original fre- quency and amplitude unchanged. Finally ,the absolute phase of the measured object is solved by using the phase or- der obtained by Gray code decoding. In addition ,the tripartite phase unwrapping method (Tri-PU) is used to elimi- nate the edge hopping error in the process of Gray code phase unwrapping. The experimental results show that the ac- curacy of the proposed method is 0. 045 mm in the reconstruction of high precision standard ball ,and the reconstruc- tion accuracy is high. The projection stripe images are reduced from four to two ,which improves the projection effi- ciency.

For the 1. 31 μm semiconductor laser working ,the far field divergence Angle is too large and the oxidation of Al component in AlGaInAs/GaAs material system leads to the degradation of device performance. Based on the InGaAsP/InP material structure ,GaAsP material is proposed to be used in the upper and lower waveguide layers of semiconductor laser ,which forms a heterojunction structure with the active region composed of InGaAsP material. Compared with the homogenous junction waveguide layer ,the heterojunction waveguide layer can achieve a larger re- fractive index difference between the active layer and the waveguide layer according to the different refractive index of different materials ,so as to achieve a better waveguide restriction. The In1-x Gax Asy P 1-y material composition in 1. 31 μm laser was calculated theoretically ,and the GaAsy P 1-y material composition and thickness were calculated the- oretically according to the refractive index and lattice constant of the material. The ALDS software is used to conduct simulation and comparative simulation analysis of the two structures ,homogeneous and heterogeneous. Finally ,GaAs0. 145P0. 855 material with a thickness of 100 nm is selected as the waveguide layer and the limiting layer respective- ly ,which effectively reduces the vertical divergence Angle of the far field and improves the circular symmetry of the spatial distribution of the laser beam.

Small target detection is a challenging research task in the field of computer vision. For the problems of small target object size ,inconspicuous features and target aggregation ,a small target detection algorithm C-SODNET based on adaptive feature fusion and task alignment is proposed. The algorithm is optimized and improved on the basis of TOOD by introducing ConvNeXt as the backbone network ,improving the feature pyramid structure by embedding CBAM attention mechanism and adaptive feature fusion module feature extraction capability of the region of interest ,while adding deformable convolution in the detection head significantly improves the detection capability for small target objects ,and finally introducing CIoU regression loss function to train the model. The experimental results show that the mAP50 of C-SODNET in VisDrone2019 small target detection dataset is 51. 2% ,which improves the accuracy rate by 9. 4% compared to the TOOD algorithm ,and the accuracy rate APs of small target objects improves by 7. 3% ,which verifies the effectiveness of the algorithm. This algorithm can provide an effective solution for small target detection ap- plications in high-altitude or long-range scenes.

A machine vision-based night-time uninterrupted monitoring technique for landslides is suggested in order to increase the capability of real- time detection of landslide dangers. By extracting the local displacement of landslide features at night ,this method tracks the regional displacement of landslides. Firstly ,the binarization method is used to extract the landslide feature point array; then ,the feature triangles of different point arrays are used to trans- form the image coordinate system to achieve the matching of points and calculate the pixel displacement. Finally ,the conversion of the pixel moving distance into the accurate distance using the principle of monocular distance measure- ment to realize slope night monitoring. The outdoor test results show that this technology can continuously monitor land- slides at night while maintaining high precision and can be applied in actual engineering.

Traditional methods for detecting methane gas leaks mainly rely on single-point measurements and infra-red thermal imaging. However ,the former method has limitations due to low measurement point density and the ease with which gas can flow ,making it difficult to determine the exact location of the leak. The latter method relies on de- tecting and imaging gas plumes based on differences in temperature between the gas cloud and the surrounding environ- ment. When the temperature difference is small ,the imaging con-trast and resolution are often low. To address these limitations ,a new methane gas leak imaging detection technology has been proposed. This approach combines TDLAS remote sensing technology with laser ac-tive imaging methods to actively detect gas leakage areas. After detecting the gas leakage areas ,an algo - rithm that combines SSIM ( Structural Similarity Index Measure) and Gaussian Mixture Background Mod-eling is used to extract the gas leakage area ,achieving imaging detection of methane gas leakage plumes. Experimental testing was conducted on methane gas leaks under different conditions ,including various leakage flow rates ,video recording lengths ,and environmental conditions. The results showed that this technology can achieve active imaging detection of methane gas leaks ,with clear imaging ,high detection rates ,good real-time performance ,and accurate localization of gas leakage points.

The traditional detection method of moving target only considers one characteristic of the image ,such as brightness or texture ,which is greatly affected by outliers ,sensitive to noise ,poor robustness ,and low precision of background recovery. In view of the above limitations ,a moving target detection method based on structural similarity ( SSIM) full reference model and robust principal component analysis ( RPCA) is proposed. In this method ,the bright- ness ,contrast and structure of the image are considered comprehensively. Instead of using the traditional background subtraction method ,the structure similarity of image pixels is taken as the metric to separate the moving objects from the background. The experimental results show that the accuracy of this method can reach 0. 95 ,and the F-measure are 0. 15 higher than the traditional moving target detection algorithm on average ,which is more advantageous than the traditional method on the whole.

The thickness of glass ,as a crucial index in industrial production ,has become a research hotspot for its rapid measurement. This paper proposes the use of laser scanning interferometry ,interpolating the sampled signals ,correcting the beat frequency nonlinearity ,and then obtaining distance information through spectral estimation. When measuring thinner glass ,the FFT algorithm is affected by the fence effect and spectral interference ,making it difficult to distinguish the frequency of the front and back surfaces of the glass in the spectrum. To address this issue ,this pa- per employs the Root-MUSIC algorithm to process the signal. Experimental results show that when the FFT algorithm fails to resolve adjacent target frequencies ,the Root-MUSIC algorithm can effectively distinguish them. The computa- tional efficiency is also much higher than the MUSIC algorithm ,and the accuracy can reach 0. 05 μm ,providing a so- lution for the problem of glass thickness measurement.

The widespread application and militarization of UAVs have brought serious harm to the national and so- cial security. Aiming at the problems of legislative bottleneck in the civil deployment of traditional anti-UAVs system equipment and the lack of simultaneous detection and recognition mechanism of multiple UAVs ,a UAVs target recogni-tion technology based on YOLOv7 was proposed. Using YOLOv7 network to identify UAV targets in high-altitude multi-scene environment : a Feature reuse based on concatenation module is introduced in YOLOv7 to solve the problems of limited feature reuse in backbone and information loss in deep network. ELAN of attention mechanism module is used to improve the ability of removing noise and suppressing irrelevant information in feature fusion. The HEAD of expan- sive convolution and residual theory is used to reduce the problem of missing detection of small target UAVs. The re- sults show that compared with the original YOLOv7 model ,the average accuracy of the improved model is increased by 2. 8% ,which solves the problem of missing detection of small targets in the original network and makes up for the shortcomings of anti-UAV in civil application.

In order to solve the problem of insufficient accuracy of image detection results obtained by traditional seam width change measurement algorithm under inclined conditions ,this paper proposes a seam width change meas- urement algorithm suitable for inclined shooting conditions. Triangular and square targets are used as left and right templates respectively. Firstly ,the appropriate size of the template is selected and the captured image is matched by template matching. Then ,the left and right templates are obtained and binarized to obtain two corner points of the left and right templates. The distance of the target is obtained by using the cross ratio invariance to obtain the width change of the seam. It is proved by experiments that this method is suitable for the measurement of joint width change under long-distance and inclined shooting. It has certain accuracy and reduces the cost of joint width change monitoring ,which provides certain value for engineering application.

In order to realize the combined detection of steel plate before and after corrosion and shock vibration state ,an ultrasonic detection system based on fiber Bragg grating ( FBG) was constructed. Firstly ,the demodulation principle of fiber grating based on tunable laser light source is analyzed theoretically. On this basis ,the variation law of Lamb wave signal before and after steel plate corrosion was studied experimentally ,and the time-frequency variation characteristics of steel plate impacted with different vertical impact intensity and the time-frequency variation charac- teristics of impact vibration signal before and after steel plate corrosion were analyzed. The research shows that the fi- ber Bragg grating ultrasonic detection system built can effectively realize the combined detection of steel plate corrosion and shock vibration signals ,and can accurately identify the corrosion of steel plates through the amplitude of the Lamb wave signal and whether there are new frequency components. There is a linear relationship between the amplitude change of the shock signal and the size of the shock. At the same time ,the frequency domain analysis proves that the frequency components of the corroded steel plate shift to the right by an average of 12 Hz.

In view of the unsatisfactory time - consuming performance of linear equation solution in the existing least squares ellipse fitting algorithm implemented on field programmable gate array( FPGA) platform ,the Cholesky decomposition algorithm was used to solve the linear equation ,and an iterative adaptive square root solution structure based on Coordinate Rotation Digital Computer ( CORDIC) algorithm was proposed by using unidirectional rotation ,merging iteration ,and adaptive adjustable iteration times to solve the square root operation in Cholesky decomposition. The experimental results show that the improved square root algorithm has a good effect on shortening the output latency of Cholesky decomposition ,and Cholesky decomposition achieves 63. 26% improvement over the existing LDLT decom- position algorithm on FPGA platform. Under the condition that the absolute error of ellipse fitting algorithm is less than 0. 1pixel ,the computer speed can be improves by more than 1 000 times based on FPGA ,so it is suitable for applica- tions with high real-time requirements.

Aiming at the problem that two-dimensional code is prone to motion blur in industrial product detec- tion ,resulting in increased recognition difficulty ,a motion blur two-dimensional code image restoration method based on improved Wiener filter is designed. In the traditional Wiener filter image restoration process ,due to the influence of the regular term K value ,resulting in the difference in the restoration effect ,this paper combines the genetic algo- rithm ,through the adaptive optimization method to achieve the estimation of K value ,the experimental results show that the improved algorithm is about 4 dB higher than the traditional algorithm restored image peak signal-to-noise ra- tio ( PSNR) ,the method can effectively restore the motion blur two-dimensional code image ,improve the efficiency of two-dimensional code recognition.

In order to solve the problem of low detection accuracy in remote sensing image target detection task due to the large number of small targets and not obvious target features ,In this paper ,an object detection algorithm based on improved YOLOv5x in remote sensing images is proposed. Firstly ,the D-SPP module is designed in the backbone network to integrate the information without deepening the network structure ,so that the characteristics of different re- ceptive fields can be effectively fused. Secondly ,SIOU~~Loss is used instead of CIOU~~Loss as the boundary frame loss function to improve the accuracy of boundary frame positioning. Finally ,add a new detection head to obtain a larger scale feature graph for target detection ,and build the smallest detection head in the network with Transformer. Experi- mental results show that the average detection accuracy of the proposed algorithm on the RSOD data set reaches 91% ,which is 5. 4% higher than that of the original YOLOv5x algorithm.

Marker points are widely used in fields such as photogrammetry and computer vision ,where the extrac- tion of image marker points is a key step for further applications in later stages. Therefore ,this paper proposes an intelligent extraction method of image marker point features based on the improved YOLOv5 model ,which has better adaptability and extraction efficiency compared with traditional algorithms. First ,a method is proposed to construct a marker point sample library based on the limit sample condition ,which can rapidly and automatically expand the marker point samples. Then ,according to the small target characteristics of marker points ,the spatial and semantic features in YOLOv5 network are fused ,and the coordinate attention mechanism is added to improve the feature extraction ability of deep learning network for marker points. The experimental results show that the correct rate of marker point extraction by the method in this paper reaches 96% ,and the average extraction time for each image is 0. 073 s. This method can provide new ideas and methods for the intelligent extraction of marker points in practical engineering.

In order to improve the effect of optical image encryption ,this paper designs an optical image encryp-tion method based on variable length key. Based on the chaos equation ,a universal algorithm for homogenization of chaotic sequences is constructed. The algorithm is used to generate homogenized pseudo-random sequences ,which are used as variable-length keys. After being modulated by a spatial light modulator ,these keys are converted into pseudo -random phase masks ( PRPM). Two converted PRPM double random phases are used to encode the initial optical image ,and combined with the deformed fractional Fourier transform ( AFrFT) ,the optical image after phase conver- sion is encoded ,Obtain encrypted optical image and complete optical image encryption. The experimental results show that this method can generate sensitive and unique keys for different optical images ,and realize the encryption of dif- ferent optical images. The encryption effect is significant ,the pixel frequency distribution of the encrypted optical im- age is uniform ,and the decrypted optical image is almost the same as the original optical image. The overall encryption and decryption effect is ideal ,which can effectively guarantee the security of the optical image ,Moreover ,the method in this paper can complete the encryption and decryption of optical images in the shortest time of 3s. The encryption and decryption efficiency is higher and the comprehensive effect is better.

In order to solve the problem of blurred face contour and single facial feature in some environment ima- ges ,a method of infrared and visible light face image registration based on machine vision technology is proposed. Ac- quire infrared and visible face images ,perform histogram equalization on them ,use Canny edge operator method to ex- tract face image contour ,and realize infrared and visible face image registration according to the gradient size and di- rection characteristics of the contour. The experimental results show that this method can effectively improve the regis- tration of infrared and visible face images. The registration value is 0. 961 when there is no occlusion ,and 0. 949 and 0. 944 when there is 10% and 20% occlusion ,respectively. This shows that the application effect of this method is sig- nificant.

A single image cannot fully describe the information of the target and has low practical application val- ue. In view of some shortcomings of the current infrared and visible image fusion methods ,such as poor fusion quality ,in order to obtain a more ideal infrared and visible image fusion effect ,an infrared and visible image fusion method based on feature similarity is proposed. First of all ,the research progress of infrared and visible image fusion is ana- lyzed ,and the limitations of various methods are pointed out. Then ,infrared and visible images are used ,and image denoising and enhancement processing are carried out. The features of infrared and visible images are extracted by con- volution neural network. Finally ,infrared and visible image fusion is carried out according to the feature similarity ,and the fusion effect of infrared and visible images is tested. The results show that ,this method improves the quality of infrared and visible image fusion ,and the fusion effect is significantly better than other infrared and visible image fu- sion methods.

The poor classification performance of spectrally similar images will increase the redundancy of spectral information and reduce the spectral detection efficiency in various fields such as ground feature exploration and military defense. In order to distinguish spectral information and spectral curves homogeneously with multiple elements ,a spec- tral similarity image classification method considering the characteristics of associated bands is proposed. The method first uses spectral matching to eliminate overexposure of white light sources in spectrally similar images. The associated bands of the optimized image are then extracted and fed into the support vector machine as clustering features. Finally ,according to the output results of the support vector machine ,the spectral similarity image classification is realized. The experimental results show that the classification results of the proposed method have high definition ,small classifi- cation errors or pixel block coloring errors ,and the classification accuracy of rectangular blocks in the same row in the confusion matrix is high.

In order to improve the virtual reconstruction effect of complex lighting images ,a virtual reconstruction method for complex lighting images based on fusion of scene depth estimation and visual communication is proposed. Aiming at the mutual interference of mixed frequency illumination at different scene depths ,a correlation matching noise reduction method is used to achieve image noise reduction processing. The median brightness value in the low brightness region of the illumination image is used as the reference value for scene depth ,and the method of global characteristics and local detail feature fitting is used to achieve scene depth detection and visual tracking fitting for complex illumination images ,The HSV spatial feature decomposition method is used to fuse the brightness channels of lighting images in different scenes ,extract detailed information such as scene object edges and textures ,and achieve virtual reconstruction of complex lighting images based on visual communication effects under scene depth detection and global contrast fusion. The simulation test results show that using this method to perform virtual reconstruction of com- plex lighting images has a good visual expression ability ,and the reconstructed image has a strong ability to display de- tails. It can accurately reconstruct hidden image information in dark areas. The peak signal to noise ratio of the two dataset images is high ,and the root mean square error is low ,respectively 45. 63 dB ,53. 21 dB ,and 0. 366 ,0. 265. Moreover ,the reconstruction time is short ,and the maximum length is only 1. 5 s ,with strong reconstruction perform-ance.

Aiming at the problems of the current infrared and visible image fusion methods such as poor fusion effect and low efficiency ,in order to obtain better infrared and visible image fusion results ,an infrared and visible image fusion method based on convolution neural network is proposed. First ,collect the infrared and visible images to be fused ,use the Retinex algorithm to enhance the image brightness and detail information ,then use the convolution neu- ral network to extract the image fusion features ,and design the infrared and visible image fusion rules ,and get the im- age fusion results according to the rules. Finally ,conduct the infrared and visible image fusion performance test on multiple data sets ,and the results show that the image fusion of convolutional neural network has good overall visual effect ,rich details ,more than 6 values of entropy and average gradient ,and the fusion time is less than 1 s. The over- all performance is better than its infrared and visible image contrast fusion method.

In order to optimize the low-light image enhancement effect ,a non-reference low-light image enhance- ment method based on visual communication is proposed. The proposed method reconstructs the low - light image through panoramic imaging technology ,uses computer principles to simulate the optical imaging system ,and introduces multifractal method and semi-soft threshold exponential attenuation method to remove the noise on the reconstructed image. The best transmittance in the process of image enhancement is calculated by the light and dark dual-channel coefficient ,and the low-light image is enhanced by this transmittance. The experimental results show that the recon- structed image is complete ,the overall quality of the image is high ,and the SIMM value is close to 1. The image de- noising is achieved without affecting the image definition. The enhancement effect on the low-light image is obvious ,avoiding the image exposure during the enhancement process ,and achieving the complete retention of image details.

Based on N×N-type sampling law of subharmonic method ,incremental sampling is introduced ,and an incremental N×N-type sampling law of subharmonic method is proposed for numerical simulation of ocean turbulence phase screen. This method uses the refractive index fluctuation spectrum proposed by Nikishov to simulate the phase of ocean turbulence ,and verifies the accuracy of the phase screen generated by the phase structure function and compen- sation effect. By changing the number of incremental regions ,sampling levels and subharmonic orders ,the influence of phase screen parameters on the accuracy of phase screen is compared and analyzed. In addition ,the proposed meth- od is compared with the subharmonic method of N× N-type sampling law ,the modified subharmonic method and the improved subharmonic method. The results show that when the subharmonic order of the incremental N×N-type sam- pling law of subharmonic method is 11 ,the compensation effect can reach 0. 999 5 ,and the simulation speed is faster with the same accuracy.

To address the problem of high false negative rate of RT - PCR in screening patients with COVID- 19 ,this paper proposes a DRPL-ViT computer - aided diagnostic network. The knowledge distillation mechanism is first introduced on the basis of Vision Transformer which enables the Transformer structure to be trained on small data sets to achieve better fitting re- sults. Then ,the dependencies between tokens can be better captured by encoding the position information of patches in a relative position encoding way that is more suitable for vision tasks. In order to focus on more local features a traditional convolution module is introduced in the Transformer Encoder module to extract local features. The experiments achieved an average classification accura- cy of 92. 11% on the four classification test sets and 97. 85% for COVID- 19. The experimental results indicate that the proposed network has a high accuracy in classifying neo - coronary pneumonia and other lung lesions ,and has some clinical application value.

To solve the UNet3+ network with depth deepening a large number of fusion redundant operation that the model training time is too long and resulting in road extraction using less problems ,the UNet3+ network improve-ment ,by cutting UNet3 + network hierarchy using Bottleneck module to replace the convolution layer in the original network ,retain the network feature fusion ability and reduce the network complexity ,and introduce hybrid attention mechanism optimization model ,build a new network model. The improvement method is compared with several typical road extraction models. The experimental results show that : (1) compared with Unet3+ network ,the proposed method improves by 4. 72% ,2. 46% ,4. 84% and 2. 01% respectively ,all better than the comparison algorithm; ( 2) com- pared with several classical feature extraction models ,the improved model has better recognition effect ,and phenoty- ping in the accuracy ,connectivity ,integrity and other aspects of road extraction.

The bus link of optical fiber communication network is easily limited by bandwidth in the application process ,resulting in poor detection accuracy of packet loss rate of optical fiber communication network link ,low infer- ence efficiency ,and unable to meet user data transmission requirements. Therefore ,a packet loss rate inference meth- od for optical fiber communication network bus link in the Internet of Things environment is proposed. According to the optical fiber communication network routing matrix ,obtain the communication status of the optical fiber network under the Internet of Things environment ,detect the weak links in the data transmission of the network bus link according to the time sequence analysis strategy ,analyze the correlation of variables between different time periods ,establish a line- ar regression model ,calculate the predicted value of the next time period ,and obtain the bus link transmission status. Calculate the spatio-temporal correlation of each line ,combine with the background traffic ,and use the least square method to deduce the link packet loss rate. The experimental results show that the fault identification rate of the pro- posed method is higher than 95% ,the error factor and relative error are below 0. 2 and 0. 02 ,and the inference time is less than 2 ms ,which provides a reference for data transmission in optical fiber communication network.

In order to analyze the impact of different threat propagation on the laser sensor network system ,and accurately and comprehensively evaluate the security of the system ,a method of laser sensor network security situation awareness under the logic regression model is proposed. The neighborhood rough set and logical regression model are combined to extract the security situation elements of the laser sensor network. By analyzing the non time series charac- teristics of different elements ,the RBF neural network is used to obtain the nonlinear mapping relationship of network security situation values ,and the laser sensor network security situation awareness model is constructed. The hybrid hi- erarchical genetic algorithm is introduced to solve the model to achieve network security situation awareness. The ex- perimental results show that the accuracy of the proposed method to obtain the security situation of the laser sensor net- work is 96. 4% ,and the error between the security situation value and the actual value is kept at - 3 ~ 3 ,which can obtain accurate changes in the security situation of the laser sensor network ,indicating that the method has certain ap- plication value.

In order to improve the security of data center optical interconnection network components and software in artificial intelligence environment ,it is necessary to build an optimized resource allocation model and propose a re- source allocation method for data center optical interconnection network based on deep neural network. The resource scheduling model of the data center optical interconnection network is constructed by using the joint optimization meth- od of user association and power spectrum allocation. The integration and clustering of different types of resources are realized by combining the QoS resource allocation of service requests for network resource granularity ,and the spatial ,temporal ,spectral and other multidimensional grid abstract model parameters of the data center optical interconnection network resources are extracted ,The deep neural network learning method is used to realize the multi-resource granu- larity fusion and convergence optimization control in the process of network resource allocation ,establish the channel model for allocating the network resources of the data center optical interconnection between users ,and realize the opti- mal allocation and balanced allocation of network resources through the transmission link balanced configuration scheme. The simulation results show that the resource allocation transmission bit rate of this method is 18 bit/s the delay is small ,the resource allocation blocking rate is low ,which is 0. 05% ,and the resource holding rate is high ,which can always be maintained at 100% ,indicating that this method has a strong ability of resource balanced alloca-tion.

For the purpose of improving the quality of optical network communication ,the paper studies the dy- namic spectrum allocation method based on crosstalk perception in time to improve the dynamic spectrum allocation effect. By calculating the crosstalk value of the optical network ,a dynamic spectrum allocation model of the optical network is established with the minimum crosstalk as the objective function ,the inter-core crosstalk generation condi- tion constraint ,and the inter-core crosstalk threshold constraint as the constraint conditions ; In quantum genetic algo- rithm ,chaos optimization algorithm is introduced to obtain improved quantum genetic algorithm to avoid falling into lo- cal optimal solution ; The improved quantum genetic algorithm is used to solve the dynamic spectrum allocation model ,and the dynamic spectrum allocation scheme of optical network with minimum inter-core crosstalk is obtained. The ex- periment shows that this method can effectively complete the dynamic allocation of optical network spectrum and reduce the inter-core crosstalk of optical network under different optical network types. Under different network loads ,the op- tical network bandwidth blocking rate of this method is 0. 17 lower ,and the spectrum utilization rate exceeds 0. 13.

The development of multi-channel optical fiber communication system not only shares the pressure of concurrent user communication requests ,but also makes the problem of signal deviation more and more serious ,lead- ing to the decline of communication reliability and the aggravation of communication delay. Aiming at the above prob- lems ,a signal bias control method for multi-channel optical fiber communication system is studied. Use the collector to obtain the channel signal and pilot channel signal of the communication system in the same time period and imple- ment filtering processing ,calculate the similarity between the two signals ,and detect whether there is signal deviation. The phase difference between the two signals is calculated by using the zero point comparison method. Through the in- troduction of correction scale factor and the adjustment of particle swarm optimization algorithm ,the PID signal devia- tion controller is optimized ,and the signal deviation control quantity of multi-channel optical fiber communication sys- tem is obtained by taking the phase difference as the input. The results show that the similarity between the three chan-nel communication signals and pilot channel signals after control is greater than that before control ,and both are grea- ter than 0. 95 ,which shows that the control method studied is effective and suppresses the signal deviation phenome- non.

Compared with 308 nm excimer laser ,355 nm pulsed solid laser has more advantages in terms of peak power ,cost and convenience. Fresh pig blood clot and fresh chicken tibia block were used as thromboid and calcified tissue to study the tissue ablation characteristics and rules of 355 nm pulsed solid laser. The experimental results show that the morphology of tissue ablation holes is good and there is no particle deposition ; The diameter of particles after ablation is 10 microns ,which is in the range of natural metabolism of human body ,and the smaller the pulse energy is in the range of 25 mJ ~ 200 mJ ,the smaller the particles after ablation ; Under perfusion conditions ,the tissue tempera- ture rise in the heat-affected zone is less than 4. 5 ℃ ,which will not cause thermal damage to the vessel wall. There- fore ,355 nm pulsed solid laser can effectively ablate thrombus and calcified tissue ,and has a good application prospect.

In the process of photonic microwave communication ,the bandwidth is narrowed due to the impact of noise ,reducing communication efficiency. In order to optimize the quality of photonic microwave communication ,a cascade filtering method for photonic microwave signals based on feedback mechanism is proposed. After removing the ambiguity of the signal using morphological erosion and dilation ,the signals with high similarity are clustered into con- tinuous regions to complete microwave signal denoising. The Kalman filter is selected to filter the microwave signal measurement noise for the first time ,and the mean filtering and dual threshold output scheme are used to filter the mi- crowave signal generated under dual photoelectric feedback for the second time. Through the cascade processing of two filters ,different clutter in the microwave signal is removed ,and the cascade filtering of narrow linewidth photonic mi- crowave signals is realized. Experimental results show that the proposed cascaded filter has a small amplitude of wave motion ,can stably output the filtering results ,and the attenuation characteristics of the cascaded filter at different tem- peratures are relatively stable.

The design effect of customized product packaging is poor due to the influence of light ,shadow and color difference. The visual image processing method of customized product packaging based on line structured light ima- ging technology. The line structured light imaging technology is used to build a visual imaging model for customized product packaging design ,analyze the difference degree of packaging image features with the change of light stripe ,and combine the depth and brightness analysis method of color space to adjust the light stability of 3D printing image in HSV color space by using the line structured light adjustment method. Through the analysis of the significance value of super pixel and the decomposition of adjacent pixel features ,Realize the integration of visual depth and brightness of customized product packaging. The experimental results show that the proposed method has better shadow and color suppression ability ,better visual image information expression ability ,the minimum mean square error is 0. 109 ,the output peak signal-to-noise ratio is 46 dB ,the peak signal-to-noise ratio is higher ,the maximum information entropy is 0. 93 ,and the SSIM value is larger when the iteration number is 10. And it ’s close to 1.

At present ,there are many errors in fault identification of photodetectors. In order to improve the effect of fault identification of photodetectors ,a fault identification method of photodetectors based on pattern recognition technology is designed. Firstly ,the state signal of the photodetector is collected ,and the features are extracted from the state signal of the photodetector. Then the principal component analysis algorithm is used to reduce the dimension of the features to obtain the optimal photodetector state identification feature. Finally ,the photodetector state feature is used as the input of the support vector machine ,and the photodetector state is used as the output of the support vector machine. The photodetector state identifier is designed through the support vector machine learning ,The experimental results show that this method can effectively identify the fault of the photoelectric detector ,the correct rate of the fault identification of the photoelectric detector is more than 90% ,and the fault identification time of the photoelectric detec- tor is controlled within 20 ms ,which provides a basis for the status analysis of the photoelectric detector.

For the laser tracking measurement task ,the result of measurement field planning is very important.The layout planning method based on laser vision sensing is designed to meet the practical application requirements.Based on laser vision sensing technology ,a laser vision sensor is designed ,which is composed of a camera ,a filter and a laser ,and is used to collect the image of the measuring field. For the collected image ,the adaptive median filter is used to de-noise it. The layout planning algorithm of the application planning engine and CAM2 is designed. In many iterations ,the particle swarm optimization algorithm is used to solve the problem ,and the layout planning of the meas- urement field is realized through this algorithm. The test results show that the layout planning results of the two stations in the design method are complementary ,indicating that the layout planning can complete the measurement task and the planning results are effective.

Aiming at improving the fault detection effect of photoelectric detection equipment ,a fault detection method of photoelectric detection equipment based on edge computing and deep learning is proposed. First ,use multi -sensor to collect the working state signal of photoelectric detection equipment ,extract features from the signal ,and then use edge computing technology to build a photoelectric detection equipment fault detection platform ,and use deep learning algorithm to model the photoelectric detection equipment fault detection samples ,and build a photoelectric de- tection equipment fault detection classifier. Finally ,simulation test results show that this method can detect various photoelectric detection equipment faults with high accuracy ,The fault detection accuracy of photoelectric detection e- quipment is more than 95% ,and the fault detection time of photoelectric detection equipment is controlled within 20 ms ,so the ideal fault detection result of photoelectric detection equipment can be obtained.

There are many heterogeneous temporal data in the electronic archives generated by laser scanning. These data have great mining value ,so it is necessary to detect them. Under this background ,a method of detecting heterogeneous temporal data of electronic archives based on laser scanning is studied. This method uses laser scanning to obtain data and generate electronic archives. For the data in electronic archives ,missing data filling and outlier data processing are implemented. The recursion rate ,Shannon entropy and skewness of electronic archive data are extrac- ted. Combined with the random forest algorithm ,the input recursion rate ,Shannon entropy and skewness are three characteristics to obtain the heterogeneous temporal data type of electronic archives. The results show that the displace- ment of 1 and 3 slope projects is slight; the displacement of two slope works is serious ; the displacement of 4 slope works is in a safe state ; the displacement of 5 slope works is relatively serious. The Kappa coefficient of the studied method reaches a relative maximum ,indicating that the detection ability of the studied method is higher.

Aiming at the problem that the current agricultural machinery controller has poor control ability on agri- cultural machinery components ,which leads to the abnormal shift and poor balance of agricultural machinery equip- ment ,the design and research of agricultural machinery controller based on laser sensor is proposed. First ,the central control module is optimized by using Nadu DPE- 10-500 laser ranging sensor as the data measurement equipment of the controller. Then set the measuring distance and angle of the sensor ,and use the circle fitting method to correct the measured data. Finally ,according to the laser measurement results ,based on the PID controller ,an adaptive neural fuzzy controller is constructed to test the control of agricultural machinery equipment. The test results show that this controller can complete the shift of mechanical equipment according to the preset requirements. The maximum position control error for the mechanical arm is only 0. 58 cm. When the intelligent balance is open ,the maximum deviation angle is only 1. 6 ° ,and the maximum deviation amplitude is only 31 cm. This greatly improves the equipment's debug- ging position accuracy and balance ,and further improves the application effect of agricultural machinery equipment.

In order to reduce the part error in the process of complex parts processing ,a laser scanning complex parts processing system based on error correction is designed. Based on laser scanning ,the overall structure of the complex parts processing system is designed. Based on the overall architecture of the system ,the laser scanning mod- ule is used to obtain the material information ,and the grid correction method is used to correct the galvanometer error in the scanning unit; The laser scanning results are transmitted to the DSP control module ,and the tool processing paths of different processing units in the processing module are determined through linear and circular interpolation op- erations. At the same time ,the feedback signals such as the positions of different servo units are compared to realize the motion control of the manipulator module. The experimental results show that the laser scanning error of the system is controlled below 0. 5 mm ,which meets the fan application standard. The maximum processing time is 37 min ,which effectively improves the application rate of parts.

Geographic mapping and effective environmental monitoring are realized through the identification of geospatial elements ,and the identification method of geospatial elements based on lidar and remote sensing data is pro- posed. The remote sensing monitoring method of laser radar to realize the geospatial characteristics ,according to the texture of geographic spatial mapping object ,shape ,spatial relationship ,using remote sensing data fusion and spatial information expression method to realize the characteristics of geospatial elements ,build geospatial element identifica- tion of remote sensing expert database. The simulation results show that the decision integration of geospatial element i- dentification is good. The method also achieves land benefits With the detection of natural feature changes ,the maxi- mum value of the recognition accuracy can reach 0. 925 ,and the recognition time is always below 2. 5 s.