Journals >Optics and Precision Engineering
Contents
2024
Volume: 32 Issue 10
15 Article(s)
Export citation format
[in Chinese]
Study on LIBS system for high-precision coal components rapid and quantitative analysis
Shulin LIU, Meng WANG, An LI, Ying ZHANG, Xiaodong LIU, and Ruibin LIU
The rapid and high-precision quantitative analysis of coal quality components is an important link for factories to efficiently utilize energy. This article quantitatively measured the industrial characteristics of fixed carbon （FixC）， ash （Ash）， volatile matter （Vdaf）， and calorific value （Q） of the collected coal samThe rapid and high-precision quantitative analysis of coal quality components is an important link for factories to efficiently utilize energy. This article quantitatively measured the industrial characteristics of fixed carbon （FixC）， ash （Ash）， volatile matter （Vdaf）， and calorific value （Q） of the collected coal samples using our laser induced breakdown technology coal quality rapid analyzer. The measured data were compared with manual measurement data in the laboratory， and it was verified that this method is accurate and fast in coal quality analysis. Through testing the stability and dynamic precision of the equipment， the results indicate that the rapid coal quality analyzer has high stability and meets the national standard requirements. When predicting coal samples with a distribution span of 5%~60% ash content， by increasing the sample size and increasing the prediction weight of high ash content coal samples in the model， the prediction accuracy of root mean square error RMSEP<1%， volatile matter and total sulfur RMSEP<1%， and calorific value RMSEP<0.18 MJ/kg can be achieved. The predicted results all meet the requirements of industrial analysis and can meet the application needs of industrial sites， with broad prospects for application in online coal quality detection..
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1470 (2024)
Durability of hydrophobic aluminum foil surfaces prepared by laser ablation imprinting
Zongbao SHEN, Kangnan MENG, Zheng SHI, Jiahui FU, and Pin LI
A laser ablation imprinting technique was proposed to prepare hydrophobic aluminum foil surfaces with multilevel microstructures. This technique used the laser to directly irradiate on the surface of the workpiece， and the primary microstructure was obtained by ablation-induced shock imprinting， and the second microstrA laser ablation imprinting technique was proposed to prepare hydrophobic aluminum foil surfaces with multilevel microstructures. This technique used the laser to directly irradiate on the surface of the workpiece， and the primary microstructure was obtained by ablation-induced shock imprinting， and the second microstructure was obtained by the flow behavior of the ablation-induced melting material. To assess the durability， acid and alkali corrosion experiments were conducted on the hydrophobic aluminum foil workpieces， as well as storage experiments under different conditions. The experimental results show that acid and alkali corrosion destroys the micro-morphology and low surface energy of the workpiece surface， thus reducing the contact angle of the corroded workpieces as the corrosion time increases. The reaction in alkali corrosion was much more intense than in acid corrosion， causing more significant damage to the surface morphology. The surface of the workpieces reached a super-hydrophilic state after 16 h of alkali corrosion （![]()
![]()
![]()
![]()
![]()
![]()
![]()
![]()
![]()
![]()
![]()
![]()
![]()
![]()
![]()
![]()
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1481 (2024)
Moisture content control strategy of magnetorheological fluid based on model-free adaptive control algorithm
Xiaoping XIAO, Liangwei LI, Jianfei ZHANG, Zisheng LI, Li CHEN, Tao ZHOU, Xing SU, and Lisheng CAI
Magnetorheological finishing is an ultra-precision machining process with stable removal efficiency and no subsurface damage. However， the water loss of magnetorheological fluid （MR fluid） in the polishing process will change the properties of polishing tools， thus affecting the stability of removal function. The existMagnetorheological finishing is an ultra-precision machining process with stable removal efficiency and no subsurface damage. However， the water loss of magnetorheological fluid （MR fluid） in the polishing process will change the properties of polishing tools， thus affecting the stability of removal function. The existing water control strategy is affected by the large time delay and time-varying disturbance of MR Fluid circulation system， which leads to the periodic fluctuation of water content and the periodic time-varying removal function， thus affecting the machining quality and accuracy. In this study， the transfer function model of MR fluid circulation system was established， and the characteristics of the system were analyzed. based on this， a Model-free adaptive control based on full form dynamic linearization （FFDL-MFAC） algorithm is designed. The algorithm can realize the parameter adaptive control of nonlinear system， effectively suppress the water fluctuation caused by time-varying disturbance and delay， and provide a simple， effective and applicable control strategy for the stable control of water content in the polishing process. The experimental results show that when FFDL-MFAC control algorithm is adopted， the Peak-valley value （PV） of the water fluctuation of MR Fluid is only 0.06%， which is reduced by 40% compared with PID. The Integral value of absolute error （IAE） was reduced by 58.1%. The stability of water content of MR fluid in polishing process is effectively improved..
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1496 (2024)
System identification and loop shaping H∞ robust control of voice coil actuator fast steering mirror
Wanchen FENG, Liang LI, and Jianqiang ZHANG
The Voice Coil Actuator-Fast Steering Mirror （VCA-FSM） driven by Voice Coil Actuator is an important servo mechanism in inter-satellite laser communications. The flexible support structure inside the FSM and the relative motion between the axes cause complex coupling properties in the system and reduce the control perfThe Voice Coil Actuator-Fast Steering Mirror （VCA-FSM） driven by Voice Coil Actuator is an important servo mechanism in inter-satellite laser communications. The flexible support structure inside the FSM and the relative motion between the axes cause complex coupling properties in the system and reduce the control performance of the system. For this problem， a two-axis loop shaping ![]()
![]()
![]()
![]()
![]()
![]()
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1511 (2024)
Study on long-term and stable imaging approach of SECCM using microfluid pressure drive
Zhiwu WANG, Jian ZHUANG, Shaohui NING, Lei CHENG, and Qiangqiang ZHENG
To enhance the imaging capability （achieve long-term and stable imaging） and avoid the evaporation and crystallization of micro-droplets at the probe tip in the scanning electrochemical cell microscopy （SECCM）， a scanning electrochemical imaging system was constructed and microfluidic pressure-driven was used in its naTo enhance the imaging capability （achieve long-term and stable imaging） and avoid the evaporation and crystallization of micro-droplets at the probe tip in the scanning electrochemical cell microscopy （SECCM）， a scanning electrochemical imaging system was constructed and microfluidic pressure-driven was used in its nanopipette probe end. The new system has advantages in terms of electrochemical imaging especially for the complex sample surface. The compensation of micro-droplet flow rate at the opening of the probe tip was studied. The research related reliability of synchronous electrochemical active and surface topography imaging also was conducted. Firstly， a SECCM imaging system with a microfluidic pressure-driven was constructed. Then， a probe detection-based numerical model that use microfluidic pressure-driven pipette was established， and the relationship between the back pressure and the fluid flow rate at the back and tip of the probe was studied. Furthermore， with the theoretical model analysis， the novel microfluidic driving method was experimentally tested for its ability to synchronously image the high-resolution topography and electrochemical activity of carbon electrode materials and aluminum alloy materials with large surface morphology characteristics. The experimental results indicate that the new method can achieve long-term and stable electrochemical imaging on the surface of hard materials （with sample surface height fluctuations greater than 20 times the probe opening diameter）. The development of the new system will provide researchers with powerful tools for studying material electrochemistry and metal material corrosion..
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1528 (2024)
[in Chinese]
Research on distributed optical fiber sensing technology for power pipe corridor breakdown monitoring
Jiaming FAN, Minnan HUANG, Yishan WANG, Mingchao YONG, and Lidong LÜ
The power pipe corridor is an important infrastructure in the city， and the monitoring and evaluation of its structure and state have attracted much attention. Aiming at the problem of external failure monitoring of power pipe corridor， this paper proposed a scheme for locating vibration events based on attenuation comThe power pipe corridor is an important infrastructure in the city， and the monitoring and evaluation of its structure and state have attracted much attention. Aiming at the problem of external failure monitoring of power pipe corridor， this paper proposed a scheme for locating vibration events based on attenuation compensation and variance threshold of optical difference time domain curve， and proved its accuracy by experiments. This scheme was based on the phase-sensitive optical time domain reflectometry system architecture. According to the phenomenon that the optical fiber vibration caused by the broken event outside the pipe corridor leaded to a sharp increase in the confusion of the optical time domain reflectometry curve at the location of the event， the optical time domain reflectometry curve obtained from different measurement sequences was differentiated， and then the difference vector data corresponding to each position of the optical fiber was evaluated and the variance threshold was set to locate the broken event. At the same time， considering the problem that the variance threshold decreased with the increase of distance due to fiber attenuation， the attenuation compensation algorithm was used to make any scattering position on the fiber correspond to the same pulse power level， so as to correct the influence of fiber attenuation on the difference threshold. A distributed fiber-optic vibration sensing system was set up in the experiment. Optical pulses with a pulse width of 30 ns and a peak power of 30 dBm were used to obtain a positioning accuracy of ±3 m in the optical fiber range of about 25 km. The external force damage event of the power pipeline gallery has the characteristics of low frequency， strong disturbance， and long duration. The proposed system scheme suppresses Rayleigh scattering fading noise through trace averaging， and combines fiber optic attenuation compensation to correct the amplitude of the scattering signal. Therefore， the phase change of the optical signal caused by vibration events is the dominant factor in trace fluctuations to ensure that external force damage events are not missed or misreported..
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1433 (2024)
Real-time video output method for underwater active polarization imaging
Houde WU, Yuzheng ZHAI, Hongchang WANG, Chiyao LI, and Ruiqi GUO
In the underwater active polarization imaging method， it is necessary to mechanically rotate the polarizer in front of the camera to obtain the orthogonal polarization state images， which makes it difficult to output the de-backscattering video in real time. The method of real-time video output method for underwater acIn the underwater active polarization imaging method， it is necessary to mechanically rotate the polarizer in front of the camera to obtain the orthogonal polarization state images， which makes it difficult to output the de-backscattering video in real time. The method of real-time video output method for underwater active polarization imaging was proposed. Firstly， the orthogonal polarization switching position was changed from the camera side to the light source side， while the polarization state of the analyzer at the camera end remained unchanged， and it was proved that this change did not affect the effect of de-backscatter. Then， the arrayed orthogonal polarization light source was designed. The light source was composed of two groups of LED lamp beads with orthogonal polarization. The two groups of LED lamp beads were controlled by an independent drive circuit to output polarized light with orthogonal polarization state， and the mechanical structure of polarization state switching was eliminated from the hardware. Third， a synchronous control device was designed to control the polarization state of the light source output and the camera to take pictures， which realized the acquisition of orthogonal polarization state image data. Finally， the pipelined active polarization image processing method was used to realize real-time video output. The experimental results show that the proposed method can output the de-backscattering video in real time， and the performance of the de-backscatter is close to that of the mechanical active polarization imaging method. The deviation degree of the de-backscattering performance of the two methods is less than 4%， and the improvement factor of the image EME value is up to 17.96. This method has the advantages of simple structure， low cost， no influence on output frame rate， and wide selectivity of the camera， and can be widely used in the underwater imaging field..
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1443 (2024)
A method for suppressing glare based on polarization feature region decomposition
Jin DUAN, Ruisen FANG, Guangqiu CHEN, Jialin WANG, Suxin MO, and Peiren GUO
In the field of sea surface detection， solar flares cause a large area of pixel saturation in photoelectric detection equipment， seriously hindering the information collection work of sea surface targets. In order to suppress solar flares， a flare suppression method based on polarization feature region decomposition waIn the field of sea surface detection， solar flares cause a large area of pixel saturation in photoelectric detection equipment， seriously hindering the information collection work of sea surface targets. In order to suppress solar flares， a flare suppression method based on polarization feature region decomposition was proposed. This method decomposed the characteristics of the flare area by analyzing the collection of polarization data by the detector. For the polarization unsaturated glare region， the polarization component image was solved using the complete polarization decomposition method to filter out some reflected light and restore the scene information of the glare obscured area. Simultaneously we set the grayscale tolerance and peel off the polarization saturated glare area. In response to the problem of data distortion in polarization saturated glare areas， FMM image restoration algorithm was adopted to diffuse water surface information from non glare areas to glare areas and reconstruct grayscale data. This article built a dazzling simulation and polarization imaging platform indoors， and conducted algorithm testing on the collected images. The experimental results show that the suppressed image reduces the number of saturated pixels by 99.98% and basically eliminates saturated pixels. At the same time， the average grayscale decreased by 59.35%. The effectiveness and feasibility of the method proposed in this article for suppressing dazzling clutter have been verified..
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1456 (2024)
[in Chinese]
Cross-level feature aggregation image enhancement with dual-path hybrid attention
Heng YUAN, Xiaoxue WANG, Tinghao YAN, and Shengchong ZHANG
To address the problems of low brightness， high noise， color deviation and loss of detail and texture in low-light images， this study proposed an image enhancement method using dual-channel hybrid attention and cross-level feature aggregation. Firstly， the Multi-scale dual-path attention residual module （MDAR） was desiTo address the problems of low brightness， high noise， color deviation and loss of detail and texture in low-light images， this study proposed an image enhancement method using dual-channel hybrid attention and cross-level feature aggregation. Firstly， the Multi-scale dual-path attention residual module （MDAR） was designed. MDAR included a Parallel multi-scale feature sampling block （PMFB） and a Dual-path hybrid attention block （DHAB）. By extracting and fusing multi-scale feature information， PMFB promoted the global representation of local features， and effectively enhanced image details. DHAB could pay more attention to image noise regions and color information， which not only alleviates the feature differences between different attention spans， but also effectively suppress noise and improve image quality. In addition， this paper designed a Cross-level feature aggregation module （CFAM）， which fuses features at different levels to make up for the differences between deep features and shallow features， strengthen the perception of shallow features， and achieve image enhancement. Experimental results indicate that the PSNR， SSIM， LPIPS and NIQE of the proposed method on the LOL dataset reached 22.347 dB， 0.850， 0.178 and 4.153 respectively and the PSNR， SSIM， LPIPS and NIQE of the proposed method on the MIT-Adobe 5K dataset reached 22.703 dB， 0.903， 0.137 and 3.822 respectively. Compared with other algorithms， the algorithm in this paper has been greatly improved， which proves the effectiveness of the proposed method..
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1538 (2024)
Optical remote sensing road extraction network based on GCN guided model viewpoint
Guanghui LIU, Zhe SHAN, Yuanhai YANG, Heng WANG, Yuebo MENG, and Shengjun XU
In optical remote sensing images， roads are easily affected by multiple factors such as obstructions， pavement materials， and surrounding environments， resulting in blurred features. However， even if existing road extraction methods enhance their feature perception capabilities， they still suffer from a large number ofIn optical remote sensing images， roads are easily affected by multiple factors such as obstructions， pavement materials， and surrounding environments， resulting in blurred features. However， even if existing road extraction methods enhance their feature perception capabilities， they still suffer from a large number of misjudgments in feature-blurred areas. To address the above issues， this paper proposed the road extraction network based on GCN guided model viewpoint （RGGVNet）. RGGVNet adopted the encoder-decoder structure and designed a GCN based viewpoint guidance module （GVPG） to repeatedly guide the model viewpoint at the connection of the encoder and decoder， thereby enhancing attention to feature blurred areas. GVPG took advantage of the fact that the GCN information propagation process had the characteristic of average feature weight， used the road salience levels in different areas as a Laplacian matrix， and participated in GCN information propagation to realize the guidance model perspective. At the same time， a dense guidance viewpoint strategy （DGVS） was proposed， which uses dense connections to connect the encoder， GVPG module， and decoder to each other to ensure effective guidance of model viewpoints while alleviating optimization difficulties. In the decoding stage， a multi-resolution feature fusion module （MRFF） was designed to minimize the information offset and loss of road features of different scales in the feature fusion and upsampling process. In two public remote sensing road datasets， the IoU of our method reached 65.84% and 69.36%， respectively， and the F1-score reached 79.40% and 81.90%， respectively. It can be seen from the quantitative and qualitative experimental results that the performance of our method is superior to other mainstream methods..
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1552 (2024)
Infrared image and visible image fusion algorithm based on secondary image decomposition
Xin MA, Chunyu YU, Yixin TONG, and Jun ZHANG
In view of the serious detail loss， the feature information of infrared image is not highlighted and the semantic information of source image is ignored in the fusion of infrared image and visible image， a fusion network of infrared image and visible image based on secondary image decomposition was proposed. The encodeIn view of the serious detail loss， the feature information of infrared image is not highlighted and the semantic information of source image is ignored in the fusion of infrared image and visible image， a fusion network of infrared image and visible image based on secondary image decomposition was proposed. The encoder was used to decompose the source image twice to extract the feature information of different scales， then the two-element attention was used to assign weights to the feature information of different scales， the global semantic branch is introduced， the pixel addition method was used as the fusion strategy， and the fusion image was reconstructed by the decoder. In the experiment， FLIR data set was selected for training， TNO and RoadScene data sets were used for testing， and eight objective evaluation parameters of image fusion were selected for comparative analysis. The image fusion experiment of TNO data set shows that in terms of information entropy， standard deviation， spatial frequency， visual fidelity， average gradient and difference correlation coefficient， SIDFuse is 12.2%， 9.0%， 90.2%， 13.9%， 85.1% ， 16.8%，6.7%，30.7% higher than DenseFuse， the classical fusion algorithm based on convolutional networks， respectively. Compared with the latest fusion network LRRNet， the average increase is 2.5%， 5.6%， 31.5%， 5.4%， 25.2% ， 17.9%，7.5%，20.7 respectively. It can be seen that the image fusion algorithm proposed in this paper has a high contrast， and can retain the detail texture of visible image and the feature information of infrared image more effectively at the same time， which has obvious advantages in similar methods..
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1567 (2024)
Underwater image enhancement synthesizing multi-scale information and attention mechanisms
Xiaohua XIA, Yuquan ZHONG, Peng HU, Yunshi YAO, Jiguang GENG, and Liangqi ZHANG
Aiming at the problems of color distortion and detail loss in underwater images due to water scattering and absorption， a generative adversarial network model integrating multi-scale information and attention mechanism was proposed to enhance underwater images. Firstly， to fully exploit and enhance both local and globaAiming at the problems of color distortion and detail loss in underwater images due to water scattering and absorption， a generative adversarial network model integrating multi-scale information and attention mechanism was proposed to enhance underwater images. Firstly， to fully exploit and enhance both local and global information of the image， local encoders and global encoders were employed to extract local and global features respectively， which were then fused to achieve complementarity. Next， a multi-scale hybrid convolution was designed to capture multi-scale information， increasing the network's adaptability to features at different scales. Subsequently， attention mechanisms were utilized to enhance the accuracy of feature extraction， emphasizing the focus on high-value features. Finally， by iteratively applying multi-scale hybrid convolution and attention mechanisms to refine features， the enhanced image was gradually up-sampled. Compared with the six classical and state-of-the-art methods， the proposed model not only achieved the best visual perception in subjective evaluations but also outperformed the six comparative methods on the entire test set in terms of four objective evaluation metrics peak signal-to-noise ratio （PSNR）， structural similarity （SSIM）， underwater image quality measurement （UIQM）， and natural image quality evaluation （NIQE） with average scores of 22.499， 0.789， 2.911， and 4.175， respectively. The improvements over the best scores among the comparative methods are 0.353， 0.002， 0.025， and 0.307， respectively. These results indicate that the proposed model not only corrects image color distortion but also performs well in restoring image details， increasing image contrast， and enhancing clarity. Therefore， it shows promising prospects for practical applications in underwater image enhancement..
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1582 (2024)
A method for dense occlusion target recognition of service robots based on improved YOLOv7
Renxiang CHEN, Tianran QIU, Lixia YANG, Tenwei YU, Fei JIA, and Cai CHEN
Aiming at the problem of poor recognition effect due to dense occlusion of the target to be recognized during visual grasping of service robots， we propose to improve the dense occlusion target recognition method for service robots with YOLOv7. First， in order to improve the problem of recognition difficulties caused bAiming at the problem of poor recognition effect due to dense occlusion of the target to be recognized during visual grasping of service robots， we propose to improve the dense occlusion target recognition method for service robots with YOLOv7. First， in order to improve the problem of recognition difficulties caused by the loss of feature information of densely occluded targets， a deep over-parameterized convolution was used to construct a deep over-parameterized high-efficiency aggregation network， and different convolution kernels were used to operate on each channel to enhance the network sensing ability， so that the network focused on the features of the target's uncovered area； second， in order to suppress the influence caused by dense occlusions and indistinguishable target boundaries on recognition， the coordinate attention mechanism was embedded into the backbone network. This enabled the network to obtain target position information and paid more attention to the important areas in the feature map， thereby enhancing the capability of the network to extract features； finally， the Ghost network was used to improve the lightweighting， reduce the number of parameters of the network model and the number of floating-point operations to realize the lightweighting， reduce the memory occupation of the model， and increase the model operation efficiency. Comparison experiments were conducted on the model in the self-constructed dataset and the public dataset respectively， and the experimental results show that the improved model achieves a mAP of 92.9% on the self-constructed dataset and 87.8% on the public dataset， which is better than the original method and the other commonly used methods. In this paper， the model reduces the memory footprint while the recognition accuracy and recognition efficiency are improved， and the overall performance is better..
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1595 (2024)
Point cloud matching algorithm based on adaptive local neighborhood conditions
Jinru LI, Jin WANG, Songtao GUO, and Hongyan SUO
To address the issues faced by traditional Iterative Closest Point （ICP） algorithms in handling complex point cloud spatial features， such as noise interference and data loss leading to slow convergence， low registration accuracy， and pool robustness， this paper proposed a point cloud matching algorithm based on adaptiTo address the issues faced by traditional Iterative Closest Point （ICP） algorithms in handling complex point cloud spatial features， such as noise interference and data loss leading to slow convergence， low registration accuracy， and pool robustness， this paper proposed a point cloud matching algorithm based on adaptive local neighborhood conditions. Initially， voxel grid filtering was used for data preprocessing， and the curvature of neighborhood surfaces was defined based on the distribution of nearby points within different radii. Considering the distribution of normal vectors and neighborhood curvature features， more accurate feature points were extracted. Subsequently， the most significantly changing curvature feature points in the neighborhood were further extracted using the least squares surface fitting method. These points were described using the Fast Point Feature Histograms （FPFH）， and similar feature point pairs were matched using a sample consensus algorithm with a set distance threshold. This calculated the key coordinate transformation parameters to complete the initial registration. Finally， a linear least squares optimization point-to-plane ICP algorithm was used to achieve more accurate registration results. Comparative experiments demonstrate that， under conditions of noise interference and data loss， the proposed method improves registration accuracy by an average of 45% and increases registration speed by 38%， compared to existing algorithms （ICP， SAC-IA+ICPK4PCS+lCP）， thus confirming its excellent robustness in handling large-volume， low-overlap point cloud registrations..
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1606 (2024)
Computer motherboard assembly defect detection using parallel feature extraction and progressive feature fusion
Junying CHEN, Zhaoyang LI, Hantao HUANG, and Xuze DONG
In view of the complex distribution of component positions， lack of prominent defect targets， and multi-scale issues in the detection of defects in computer motherboard assembly， this paper proposed an end-to-end defect detection algorithm based on parallel feature extraction and cross-attention progressive feature fusIn view of the complex distribution of component positions， lack of prominent defect targets， and multi-scale issues in the detection of defects in computer motherboard assembly， this paper proposed an end-to-end defect detection algorithm based on parallel feature extraction and cross-attention progressive feature fusion. Firstly， a parallel residual feature extraction network was proposed by combining partial convolution and visual Transformer. The low computational complexity of partial convolution was utilized to extract local features， while the long-distance modeling ability of visual Transformer was utilized to expand the receptive field of the model and enhance the feature extraction ability of the network. Secondly， the cross-attention mechanism was introduced to progressively fuse multi-scale features， and a multi-scale cross-attention progressive feature fusion network was constructed to enhance the feature fusion ability of the detection model. The experimental results on the public dataset show that the mean average accuracy （mAP） of the algorithm reaches 94.63%， which is 4.62% higher than the baseline model YOLOv5 and is superior to several other advanced models. The detection speed reaches 25 FPS， achieving a good balance between detection accuracy and speed. It provides a fast and effective method for the automation and intelligence of surface assembly defect detection on computer motherboards in the actual industrial environment..
Optics and Precision Engineering
- Publication Date: May. 25, 2024
- Vol. 32, Issue 10, 1622 (2024)
© Copyright 2018-2021 | Chinese Laser Press. All Rights Reserved 沪ICP备15018463号-20