Journals >Laser & Optoelectronics Progress
Contents
2023
Volume: 60 Issue 24
29 Article(s)
Export citation format
Atmospheric Optics and Oceanic Optics
Feature-Enhanced Cloud Image Prediction Algorithm Based on Spatio-Temporal Attention Gated Recurrent Unit
Xiuzai Zhang, Jingxuan Li, Changjun Yang, and Xuan Feng
Changes in various weather phenomena are accompanied by the movement of clouds. Continuous satellite cloud images obtained by meteorological satellites contain considerable spatio-temporal sequence information; that is, continuous satellite cloud images have significant time sequence characteristics, which can be used Changes in various weather phenomena are accompanied by the movement of clouds. Continuous satellite cloud images obtained by meteorological satellites contain considerable spatio-temporal sequence information; that is, continuous satellite cloud images have significant time sequence characteristics, which can be used as basic information for cloud image prediction. Cloud image prediction is essentially a video prediction problem in which the spatio-temporal sequence characteristics of cloud images are analyzed and processed. To accurately predict the change in the position of clouds, by focusing on the unstable and nonlinear motion characteristics of clouds, a video prediction algorithm called SmartCrevNet for cloud image prediction based on the CrevNet video prediction model is proposed. In this algorithm, a spatiotemporal attention gated recurrent unit (STA-GRU), along with the lightweight attention module spatial group-wise enhance (SGE), is introduced into the original two-way autoencoder module of CrevNet to enhance the ability of feature extraction without increasing the amount of calculation. The algorithm was tested on the public dataset MovingMNIST and the FY-4A satellite cloud image dataset. The results show that, on the FY-4A satellite cloud image dataset and MovingMNIST dataset, the mean square error (MSE) of SmartCrevNet is, respectively, 7.3% and 6.1% lower than that of CrevNet, and the structural similarity (SSIM) is increased by 7.9% and 1.2%, respectively. The prediction effect is better than that of CrevNet and traditional video prediction algorithms..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2401001 (2023)
China's Top 10 Optical Breakthroughs Special Column
China's Top 10 Optical Breakthroughs: Research Progress of Photonic Skyrmion
Min Lin, Luping Du, and Xiaocong Yuan
Recently, photonic skyrmion as a topological nontrivial structure has attracted widely research attention. Due to their ultracompact size, high stability, and diversity of topologies, photonic skyrmions have potential value in applications of high-resolution polarization imaging, high-density optical information storagRecently, photonic skyrmion as a topological nontrivial structure has attracted widely research attention. Due to their ultracompact size, high stability, and diversity of topologies, photonic skyrmions have potential value in applications of high-resolution polarization imaging, high-density optical information storage, and high-precision displacement sensing. In this paper, the fundamental mechanisms and the excitation and detection methods of photonic skyrmion are introduced, and the domestic and foreign state-of-the-art studies on the photonic skyrmion in different optical systems are summarized. In view of the deep-subwavelength features of the photonic skyrmion, the recent research progress of the applications of photonic skyrmions is also reviewed, and its future development is analyzed and prospected..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2400001 (2023)
Image Processing
Non-Intrusive Load Identification Method Based on Gramian Angular Difference Field Image Coding
Ming Fu, and Bin Duan
Non-intrusive load monitoring, as an essential means for fine-grained management of household electricity consumption, plays a significant role in promoting energy conservation and emission reduction for achieving the dual-carbon goal. However, it is challenging to achieve high-precision load identification using a sinNon-intrusive load monitoring, as an essential means for fine-grained management of household electricity consumption, plays a significant role in promoting energy conservation and emission reduction for achieving the dual-carbon goal. However, it is challenging to achieve high-precision load identification using a single voltage-current trajectory image. Therefore, a non-intrusive load identification method based on the fusion of Gramian angular difference field (GADF) image coding is proposed. First, the high-frequency steady-state data collected by the device are preprocessed to obtain a complete base-wave period current and voltage signal. Then, the one-dimensional voltage and current signals are encoded separately using the GADF to generate the corresponding two-dimensional feature images, and load identification is performed via superimposed fusion input to a neural network based on a convolutional block attention module. The public datasets PLAID and WHITED are used for testing experiments to verify the effectiveness of the proposed method. The results indicate that the method has a high recognition accuracy, with average accuracies of 99.45% and 99.24% for the PLAID and WHITED datasets, respectively..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2410001 (2023)
Infrared and Visible Image Fusion Based on Separate Expression of Mutual Information Features
Hui Wang, Xiaoqing Luo, and Zhancheng Zhang
To solve the challenges associated with the inadequate separation of source image features, low interpretability, and difficulty of designing accurate fusion rules, this paper proposes an infrared (IR) and visible image fusion method based on mutual information feature separation and representation, which effectively sTo solve the challenges associated with the inadequate separation of source image features, low interpretability, and difficulty of designing accurate fusion rules, this paper proposes an infrared (IR) and visible image fusion method based on mutual information feature separation and representation, which effectively separates features while preserving the typical information of the source image. First, a mutual information constrained coding network is used to extract the features, maximize the mutual information between the source image and features to retain the feature representation of the source image, and minimize the mutual information of private and public features to achieve separation and representation. In addition, the loss function adopts a soft weighted intensity loss to balance the distribution of IR and visible features. Objective and subjective evaluation results of comparison experiments indicate that the proposed method can effectively fuse important information regarding IR and visible images and has good visual perception..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2410002 (2023)
Object Detection via Multimodal Adaptive Feature Fusion
Xiaoqiang Gao, Kan Chang, Mingyang Ling, and Mengyu Yin
With the advancement of deep learning, object detection methods based on convolutional neural networks (CNNs) have achieved tremendous success. Existing CNN-based object detection models typically employ single-modal RGB images for training and testing; however, their detection performance is significantly degraded in With the advancement of deep learning, object detection methods based on convolutional neural networks (CNNs) have achieved tremendous success. Existing CNN-based object detection models typically employ single-modal RGB images for training and testing; however, their detection performance is significantly degraded in low-light conditions. To address this issue, a multimodal object detection network model built on YOLOv5 is proposed, which integrates RGB and thermal infrared imagery to fully exploit the information provided by the fusion of multi-modal features, increasing the object detection accuracy. To achieve effective fusion of multimodal feature information, a multimodal adaptive feature fusion (MAFF) module is introduced. It facilitated multimodal feature fusion by adaptively selecting diverse modal features and exploiting the complementary information between modalities. The experimental results indicate the efficacy of the proposed algorithm for seamlessly merging features from distinct modalities, which significantly increases the detection accuracy..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2410003 (2023)
Infrared Vehicle Detection Algorithm Based on Improved Shuffle-RetinaNet
Xiaochang Fan, Yu Liang, and Wei Zhang
In view of the low detection accuracy and high complexity of current multi-scale vehicle detection algorithms in infrared scenes, an infrared vehicle detection algorithm based on Shuffle-RetinaNet is proposed. On the basis of RetinaNet, the algorithm uses ShuffleNetV2 as the feature extraction network. A dual-branch atIn view of the low detection accuracy and high complexity of current multi-scale vehicle detection algorithms in infrared scenes, an infrared vehicle detection algorithm based on Shuffle-RetinaNet is proposed. On the basis of RetinaNet, the algorithm uses ShuffleNetV2 as the feature extraction network. A dual-branch attention module channel attention module is proposed, which adopts the dual-branch structure and adaptive fusion and enhances the ability to extract the key features of the target in infrared images. To optimize the feature fusion, the algorithm integrates cross-scale connection and fast normalized fusion in some feature layers to enhance the multi-scale feature expression. The calibration factor is set to enhance the task interaction of classification and regression, and the accuracy of target classification and locating is increased. A series of experiments are conducted on a self-built infrared vehicle dataset to verify the effectiveness of the proposed algorithm. The detection accuracy of this algorithm for the self-built vehicle dataset is 92.9%, the number of parameters is 11.74×106, and the number of floating-point operations is 24.35×109. The algorithm exhibits better detection performance on the public dataset FLIR ADAS. Experimental results indicate that the algorithm has advantages in detection accuracy and model complexity, giving it good application value in multi-scale vehicle detection tasks in infrared scenes..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2410006 (2023)
A Low-Light Image Enhancement Method Combined with Generative Adversarial Networks in Nonsubsampled Shearlet Transform Domain
Wenling Shi, Yipeng Liao, Zhimeng Xu, Xin Yan, and Kunhua Zhu
Low illumination image has a number of issues, such as low recognition, low brightness, low resolution, low signal-to-noise ratio and blurred details. Therefore, a low-light image enhancement method combined with generative adversarial networks (GAN) in nonsubsampled shearlet transform (NSST) domain is proposed. First,Low illumination image has a number of issues, such as low recognition, low brightness, low resolution, low signal-to-noise ratio and blurred details. Therefore, a low-light image enhancement method combined with generative adversarial networks (GAN) in nonsubsampled shearlet transform (NSST) domain is proposed. First, low-light image and normal light image datasets are collected, the images are processed by RGB to HSV spatial transformation, the Hue and the Saturation components are unchanged, the Value components are decomposed at multiple scales by NSST, and the decomposed low-pass subband images are used to construct training set. Second, a low-frequency subband image enhancement model based on GAN is constructed, and the low-frequency subband image training set is used to train the model. Then, the low-illumination image to be processed is decomposed by NSST, the trained model is used to enhance the low-frequency subband image, the scale correlation coefficient is used to remove noise for each high-frequency direction subband, and the edge coefficient is enhanced by the nonlinear gain function. Finally, NSST reconstruction is performed on the low-frequency and high-frequency subband images after enhanced processing, and the reconstructed images are restored to RGB space. In terms of low-light image enhancement, compared to common methods, the results obtained by the proposed method show an average improvement of 3.89% in structural similarity and an average reduction of 1.03% in mean squared error, and when the noisy images are enhanced, the peak signal to noise ratio and continuous edge pixel ratio remain above 21 dB and 88%, respectively. The experimental results show that both visual effect and objective evaluation index of image quality of the proposed method are greatly improved compared to the common methods, which can effectively improve the low-quality problem of low-light images, and lay the foundation for the subsequent image processing analysis..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2410007 (2023)
High Dynamic Infrared Image Compression and Enhancement Algorithm Based on Side Window Filtering
Xianzhen Sang, Hongtai Zhu, Hu Cheng, Min Li, Kai Hu, Jun Tang, Mingdong Hao, and Zheng Yuan
Data collected via infrared thermal imaging systems are primarily in high dynamic range. Thus, research on dynamic range compression and detail enhancement technology is crucial to achieve visualization of high dynamic infrared images. This paper addresses the challenges of gradient reversal artifacts, low contrast detData collected via infrared thermal imaging systems are primarily in high dynamic range. Thus, research on dynamic range compression and detail enhancement technology is crucial to achieve visualization of high dynamic infrared images. This paper addresses the challenges of gradient reversal artifacts, low contrast detail loss, and background noise over enhancement in traditional methods. In this paper, we propose a high dynamic infrared image compression and enhancement method based on side window filtering. First, side window filtering is used to decompose the original infrared image into basic and detail components. Then, an adaptive threshold platform histogram algorithm is designed based on the grayscale distribution of the basic component in order to compress the basic component. The detail component is enhanced using the adaptive gain coefficient generated via the weight distribution characteristics of the bilateral filter core. Finally, the basic and detail components are weighed and fused and quantified to an 8-bit dynamic range. According to experimental results, compared with classic compression enhancement methods, the proposed method has a superior edge preservation effect on strong edges, can effectively avoid gradient inversion artifacts and halo problems, and has richer detail information, better background noise suppression effect, and stronger adaptability to different scenes..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2410009 (2023)
Fracture Zone Extraction Method Based on Three-Dimensional Convolutional Neural Network Combined with PointSIFT
Hao Wang, Dongmei Song, Bin Wang, and Song Dai
This paper presents a fracture zone extraction method for complex terrain areas using a combination of a three-dimensional convolutional neural network (3D-CNN) and PointSIFT. The PointSIFT module encodes spatial orientation information of the original point cloud data to aggregate point cloud features, resulting in reThis paper presents a fracture zone extraction method for complex terrain areas using a combination of a three-dimensional convolutional neural network (3D-CNN) and PointSIFT. The PointSIFT module encodes spatial orientation information of the original point cloud data to aggregate point cloud features, resulting in reconstructed point cloud data with different scale features. Subsequently, a 3D-CNN model is developed, with a 3D convolutional module serving as the primary component, to extract deep-level features from the reconstructed point cloud data. The extracted point cloud features are then fed into a fully connected layer for the categorization of the point clouds, addressing the challenge associated with fracture zone extraction. Comparative evaluations with the tensor decomposition method and deep neural network method are performed on two datasets. The results demonstrate that the proposed fracture zone extraction method achieves a lower classification error, thus confirming the superiority of the method in effectively extracting fracture zones from point cloud data..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2410011 (2023)
Curved Texture Flattening Algorithm Based on the Light Field Camera
Shengnan Qin, and Yanting Lu
For texture patterns distributed on the curved surfaces of objects, it is often necessary to extract and flatten the texture patterns from the curved surfaces for the purpose of comprehensive display and subsequent usage. Therefore, we suggest a curved texture flattening scheme based on the light field camera, especialFor texture patterns distributed on the curved surfaces of objects, it is often necessary to extract and flatten the texture patterns from the curved surfaces for the purpose of comprehensive display and subsequent usage. Therefore, we suggest a curved texture flattening scheme based on the light field camera, especially the focused light field camera, which can provide a high-resolution texture image and corresponding depth map without additional registrations. A curved texture flattening algorithm is designed for this scheme. The algorithm divides the curved texture image into multiple overlapping local texture images, corrects the local texture distortion based on the normal vector of the fitted plane for each local texture image, and finally stitches the corrected local texture images into a completely flattened texture image. Simulated and real experiments reveal that the proposed curved texture flattening algorithm can effectively flatten a variety of texture patterns distributed on different curved surfaces, and the algorithm has certain robustness on the different image quality of texture images and the errors of the depth measurements..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2410012 (2023)
Imaging Systems
Soft Histogram of Gradients Loss: A loss Function for Optimization of the Image Fusion Networks
Yuxin Long, Wenjie Lai, Huaiyuan Zhang, Hongbo Zhang, Chengshi Li, and Ziji Liu
Image fusion methods based on deep learning have achieved excellent image fusion performance and have been widely used in biometric recognition, automatic driving and target tracking. However, it is still challenging to extract important texture details and preserve information of images. Therefore, a loss function forImage fusion methods based on deep learning have achieved excellent image fusion performance and have been widely used in biometric recognition, automatic driving and target tracking. However, it is still challenging to extract important texture details and preserve information of images. Therefore, a loss function for infrared and visible image fusion networks is presented. We employ the histogram of oriented gradient (HOG) to calculate the loss function. HOG feature can reflect the direction and magnitude of local gradient in the image, and using HOG feature as the loss function can improve the ability of the network to extract image details. We combine HOG loss with multi-scale structural similarity loss, and train NestFuse, Res2Fusion and UNFusion infrared and visible image fusion networks with the designed loss function. On the TNO dataset, our model increases the standard deviation (SD) of fused images by 2.1476%, 1.2273% and 1.4444% respectively, and increases the visual information fiedity (VIF) of fused images by 1.6529%, 1.4936% and 1.2902% respectively. On the RoadScene dataset, our model increases the SD of the fused images by 1.0083%, 1.1669% and 0.7214% respectively, and increases the VIF of the fused images by 1.8093%, 1.8063% and 1.0406% respectively. The experimental results show that the proposed loss function can extract more effective information from the source image..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2411001 (2023)
Instrumentation, Measurement and Metrology
Oil Spill Detection Algorithm of a Fully Polarimetric SAR Based on Dual-EndNet
Dongmei Song, Mingyue Wang, Chengcong Hu, Jie Zhang, Bin Wang, Shanwei Liu, Dawei Wang, and Bin Liu
Marine oil spill accidents not only result in huge property and economic losses but also adversely affect the marine ecosystem. The polarimetric synthetic aperture radar (PolSAR) is widely used for marine oil spill detection because it can record the backscattering information of ground objects comprehensively through Marine oil spill accidents not only result in huge property and economic losses but also adversely affect the marine ecosystem. The polarimetric synthetic aperture radar (PolSAR) is widely used for marine oil spill detection because it can record the backscattering information of ground objects comprehensively through various polarization channels. To detect offshore oil spills more accurately, this study proposes a PolSAR marine oil spill detection algorithm based on a Dual Encoder-Decoder Net (Dual-EndNet). First, the 30 polarimetric features commonly used for oil spill detection were extracted from the data, and the top 10 features with high importance for oil spill detection were selected by a random forest algorithm. Next, using the encoder-decoder as the basic framework, the two branches were designed to input the PauliRGB images and the selected 10 polarimetric feature images, respectively. These were used to extract the spatial information and polarization information from the PolSAR images of the oil spill. Then, the two branches of information are merged to improve the network performance. Experiments conducted on two Radarsat-2 fully PolSAR oil spill datasets show that the proposed method has a strong oil spill detection capability, and can effectively distinguish different types of oil films, including mineral oil, biogenic film, and emulsions..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2412002 (2023)
An Improved YOLOv5 Algorithm for Steel Surface Defect Detection
Shaoxiong Li, Zaifeng Shi, Fanning Kong, Ruoqi Wang, and Tao Luo
Scale of steel surface defects is different, but existing detection algorithms have poor multi-scale feature processing ability and low accuracy. Therefore, an improved YOLOv5 algorithm for steel surface defect detection is proposed. First, receptive field modules are added after the feature output layer of the backbonScale of steel surface defects is different, but existing detection algorithms have poor multi-scale feature processing ability and low accuracy. Therefore, an improved YOLOv5 algorithm for steel surface defect detection is proposed. First, receptive field modules are added after the feature output layer of the backbone to enhance the discrimination and robustness of the features which can better perceive the feature information of different scales. Then, aligned feature aggregation modules are used to replace the traditional feature fusion structure to solve the feature misalignment problem in the fusion process of high and low resolution feature maps. Finally, decoupled heads with efficient channel attention mechanisms are used to output the detection results. The attention mechanism can adaptively calibrate the channel response, and the decoupled heads enable classification and regression tasks to be performed independently. The experimental results on NEU-DET dataset show that the mean average precision of the proposed method is 80.51%, which is 4.48% higher than that of the benchmark model, and the detection speed is 31.96 frame/s. Compared with other mainstream object detection algorithms, the proposed algorithm has higher accuracy while maintaining certain detection speed, enabling efficient steel surface defect detection..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2412003 (2023)
Infrared Image Fault Detection of Photovoltaic Modules Based on Residual Photovoltaic Network
Mingzheng Sun, and Hao Li
At present, infrared image data of photovoltaic modules obtained by unmanned aerial vehicle is more and more used in the fault detection of photovoltaic module. However, due to the high similarity of various samples of photovoltaic module infrared image data, the existing deep learning model has a low ability to extracAt present, infrared image data of photovoltaic modules obtained by unmanned aerial vehicle is more and more used in the fault detection of photovoltaic module. However, due to the high similarity of various samples of photovoltaic module infrared image data, the existing deep learning model has a low ability to extract photovoltaic module infrared image features, resulting in low detection accuracy of photovoltaic module multi-fault types. To solve the above problems, a residual photovoltaic network (ResPNet) model is constructed based on the residual network (ResNet) model for infrared image fault detection of photovoltaic modules. On the basis of ResNet model, ResPNet adds the underlying feature information enhancement module, multi-scale feature information enhancement module and global feature information enhancement module to improve the infrared image feature extraction ability of photovoltaic modules. Experiments are conducted on Infrared Solar Modules, a disclosed infrared image dataset of photovoltaic modules. The ResPNet model achieves an infrared image classification accuracy of 84.6% for 12 types of photovoltaic modules, which is better than not only ResNet-50 model, but also other infrared image classification models. Through cascading several ResPNet models, the highest known infrared image classification detection accuracy of 12 types of photovoltaic modules in this dataset is achieved at 85.9%..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2412005 (2023)
Polished Surface Defect Detection Based on Intelligent Surface Analysis
Zihao Li, Fengzhou Fang, Zhonghe Ren, and Gaofeng Hou
Surface quality of the workpiece is critical for part reliability, quality and service life. Although various vision-based target detection frameworks have been widely applied to industrial surface defect detection scenarios, surface defect detection of ultra-precise machining workpieces is still challenging due to theSurface quality of the workpiece is critical for part reliability, quality and service life. Although various vision-based target detection frameworks have been widely applied to industrial surface defect detection scenarios, surface defect detection of ultra-precise machining workpieces is still challenging due to the influence of face shape and the confounding nature between defects. Therefore, we propose a frequency-embedded two-branch parametric prediction network to predict the filtering parameters and filter out the profile information to make the defect features more significant. After pre-processing based on intelligent type surface analysis, a cascaded regional neural network-based perceptual field enhancement defect detection network is proposed. It replaces the deformable convolution intervals into the convolution module of the EfficientNet, which effectively improves the feature extraction capability of the backbone network. Then, the feature map is reselected to form a new feature pyramid network to improve the efficiency and further improve the network performance. In addition, the filter parameter dataset ultra precision polishing (UPP-CLS) with filter parameter labelling information and the defect detection dataset UPP-DET with defect category and location are constructed. The model achieves 85.36% accuracy on UPP-CLS, which is 3 to 5 percentage points higher than that of the existing networks, and 0.862 average precision on UPP-DET, which is 5.3%?7.8% higher than that of the existing networks. The overall performance of the model is better than the mainstream network architecture. The source code and dataset will be available at https://gitee.com/zihaodl/detect_app ..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2412006 (2023)
Detection of Surface Defects in Lightweight Insulators Using Improved YOLOv5
Yu Guo, Meiling Ma, and Dalin Li
Herein, an improved insulator defect-detection algorithm, YOLOv5, is proposed to overcome the shortcomings, including inconspicuous target features and poor detection of small targets when detecting trapped insulators using unmanned aerial vehicles, which cannot satisfy both detection speed and accuracy. First, ConvNeXHerein, an improved insulator defect-detection algorithm, YOLOv5, is proposed to overcome the shortcomings, including inconspicuous target features and poor detection of small targets when detecting trapped insulators using unmanned aerial vehicles, which cannot satisfy both detection speed and accuracy. First, ConvNeXt is applied to the YOLOv5 reference network to improve its ability to extract the features of obscure targets. Moreover, a coordinate attention mechanism is introduced into the reference network to improve its detection accuracy with respect to small targets in an image. Then, the improved model is pruned to eliminate its redundant channels, thus reducing the number of model parameters and making the model more lightweight. The experimental results show that the improved model achieves an average detection accuracy of 93.84% with respect to the insulator-defect dataset IDID, which is 3.4 percentage points higher than the accuracy achieved by the original algorithm. Moreover, the highest detection rate achieved by the proposed algorithm is 166 frame/s, which is 69.4% higher than that achieved by the original algorithm. These results prove that the improved algorithm meets the requirements of real-time transmission-line detection..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2412007 (2023)
Machine Vision
Wheel Tread Anomaly Detection Based on Attentional Reverse Knowledge Distillation
Rongrong Qin, xiaorong Gao, Lin Luo, and Jinlong Li
Wheels are an essential part of railway trains; thus, defects on the wheel tread present serious risk regarding the safety of railway trains. Due to the limited samples of wheel tread defects in practice, the corresponding supervised detection model is insufficient. To solve this problem, an unsupervised knowledge distWheels are an essential part of railway trains; thus, defects on the wheel tread present serious risk regarding the safety of railway trains. Due to the limited samples of wheel tread defects in practice, the corresponding supervised detection model is insufficient. To solve this problem, an unsupervised knowledge distillation anomaly detection model is proposed to detect wheel tread anomalies. Accordingly, UNet is employed to segment the tread region and reduce the influence of non-tread regions on the anomaly detection model. An attention mechanism is then added after the multiscale feature fusion to improve the ability of the student network to reconstruct normal features in the reverse knowledge distillation structure, as well as enhance the reconstruction of normal features. From the experimental results, the improved model achieves the performance indexes of 93.8% area under receiver operating characteristic curve, 82.3% precision, 95.4% recall, and 87.0% accuracy considering the railway wheel tread dataset. Compared with the original model, the detection performance of the model is improved..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2415002 (2023)
Point Cloud Analysis Method Based on Spatial Feature Attention Mechanism
Yanlin Qu, Yue Wang, Qian Zhang, and Shaokun Han
To address the limitations of existing point-based networks, which treat all points with equal emphasis, thereby overlooking crucial features, this paper introduces an attention mechanism to lidar point cloud processing. This mechanism, referred to as the CSA module, integrates the channel attention and spatial attentiTo address the limitations of existing point-based networks, which treat all points with equal emphasis, thereby overlooking crucial features, this paper introduces an attention mechanism to lidar point cloud processing. This mechanism, referred to as the CSA module, integrates the channel attention and spatial attention elements. In a data-driven approach, the two proposed modules autonomously learn the importance of different feature channel information and spatial location information, thereby enhancing the performance of the network on point cloud classification and segmentation tasks. This paper introduces the two modules stated above in a point-based network and proposes a CSA-PointNet++ architecture. The results reveal that the proposed method achieves an accuracy of 93.20% for classification experiments on the ModelNet40 dataset and a mean intersection over union (mIoU) of 82.62% for part segmentation experiments on the ShapeNetPart dataset. This performance is better than that of other comparative methods, indicating the effectiveness of the proposed network. Moreover, classification experiments of the proposed method on a real-world self-constructed dataset yield an accuracy of 92.14%, demonstrating the excellent generalization capability of the proposed network on real-world data..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2415003 (2023)
Optical Design and Fabrication
Lithography Hotspot Detection Based on Improved Yolov5s
Qingyue Wu, Jiamin Liu, Song Zhang, Hao Jiang, and Shiyuan Liu
Lithography hotspot detection plays a critical role in realizing the manufacturability design of integrated circuits (IC) and ensuring the final yield of IC chips. Considering that conventional lithography hotspot detection methods based on deep learning are challenging to meet the inspection precision requirement of aLithography hotspot detection plays a critical role in realizing the manufacturability design of integrated circuits (IC) and ensuring the final yield of IC chips. Considering that conventional lithography hotspot detection methods based on deep learning are challenging to meet the inspection precision requirement of advanced IC manufacturing, we propose a detection algorithm based on improved Yolov5s for the precise detection of hotspot defects in the lithography layout. In the algorithm, a coordinate attention mechanism is introduced into the backbone network, which can improve the attention of the Yolov5s model to the patterned area in the layout. Thereby, the performance of the lithography hotspots based on the Yolov5s detection algorithm can be greatly promoted. Meanwhile, the Sigmoid linear unit activation function is used to improve the nonlinear expression of the entire neural network, and the Scylla intersection over union loss function is adopted to realize the quantitative evaluation of the bounding box regression loss more quickly, which can enhance the convergence speed and accuracy of the algorithm. Using the ICCAD (The International Conference on Computer-Aided Design) 2012 contest benchmark and the optical proximity correction optimized lithography patterns as the dataset, performance test experiments are carried out to verify the excellent detection accuracy of the proposed algorithm. The experimental results indicate that the mean precision, mean recall, mean F1-score, and mean average precision of the algorithm reach 97.7%, 98.0%, 97.8%, and 98.4%, respectively, which are significantly better than those of other hotspot detection algorithms and show its good application prospects..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2422001 (2023)
Optical Devices
Developing Dynamic Compensation System for Eye Defocus Based on Image Processing
Zhongxiang Zhang, Chongxiang Xu, Weijiang Yan, Yufeng Su, and Zhigang Jia
The commonly used correction methods of ametropia are all focused on correcting a certain static value of human-eye aberration; however, these methods cannot effectively correct the dynamic changes in human-eye aberration. To solve this problem, a dynamic compensation system is developed to correct the eye's defocuThe commonly used correction methods of ametropia are all focused on correcting a certain static value of human-eye aberration; however, these methods cannot effectively correct the dynamic changes in human-eye aberration. To solve this problem, a dynamic compensation system is developed to correct the eye's defocusing of humans based on image processing. The system consists of a dynamic measurement system for pupil size composed of an infrared camera and an image processing program, and a dynamic correction system for refractive power composed of a transmission deformable mirror and a control program as the core. The system prototype is built on the optical experimental platform, where the corresponding experiments are performed. The results show that the system can assess the accurate measurement of the pupil size and ensure an accurate, fast, and smooth correction of the corresponding defocus aberration under different light intensities, thereby preliminarily confirming the feasibility and effectiveness of the proposed method and system..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2423001 (2023)
Remote Sensing and Sensors
Loop Closure Detection Algorithm Based on Single Frame-Submap Descriptor Matching
Lianxin Dong, Kangnian Wang, and Zhanhua Huang
Small viewing field of the solid-state LiDAR makes loop closure detection difficult. Therefore, a loop closure detection method based on single frame-submap descriptor matching is proposed. First, to obtain a submap with the pose provided by the front-end odometry, several frames of point clouds are spliced, then the dSmall viewing field of the solid-state LiDAR makes loop closure detection difficult. Therefore, a loop closure detection method based on single frame-submap descriptor matching is proposed. First, to obtain a submap with the pose provided by the front-end odometry, several frames of point clouds are spliced, then the descriptors of the submap are obtained, and their positions are added to the KD tree. Second, for each current frame, the KD tree is used to search candidate submaps, and the descriptor is obtained after projecting to the submap coordinate system according to the pose getting from odometry, so as to realize the rotation and translation invariance of the descriptor. Then, the binary descriptor is used for alignment, and the mask method is used to calculate the similarity between the current frame descriptor and the submap descriptor. Finally, the CFB-ICP algorithm is used to register the qualified loop closure pairs to obtain the loop closure factors, and factor graph optimization is carried out. Experimental tests are carried out in open source data sets and real outdoor environments. The results show that the algorithm can reduce the cumulative error under long-distance operations and improve the accuracy of positioning and mapping under the premise of satisfying real-time performance..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2428002 (2023)
LiDAR Point Cloud Correction and Location Based on Multisensor Fusion
Wenhao Pu, Xixiang Liu, Hao Chen, Hao Xu, and Ye Liu
It is difficult to solve motion distortion and poor positioning accuracy caused by point cloud distortion and error accumulation in a LiDAR moving scene using a single sensor. To address this problem, a LiDAR point cloud distortion correction and positioning method that combines inertial measurement unit data and wheelIt is difficult to solve motion distortion and poor positioning accuracy caused by point cloud distortion and error accumulation in a LiDAR moving scene using a single sensor. To address this problem, a LiDAR point cloud distortion correction and positioning method that combines inertial measurement unit data and wheel tachometer data is proposed. First, the data of the inertial measurement unit and the wheel tachometer are preprocessed by an integration method based on the time of the LiDAR data. Next, the fusion data and the LiDAR point cloud data are fused to correct the position and pose of the laser point cloud distorted by motion. Finally, the linear interpolation method is used to ensure the time synchronization and availability of data between sensors and ultimately improve the positioning accuracy of the odometer; the calculated pose was used as the optimal initial value of the odometer iteration. The experimental results show that compared with the traditional method that does not use multisensor fusion (LOAM and F-LOAM), the proposed method's root mean square error of positioning on the open data set experiment is reduced by 81.11% and 21.54%, respectively, the root mean square error of positioning of the proposed method on the self-testing data concentration period is reduced by 52.76% and 24.29%, respectively..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2428003 (2023)
Imaging Simulation Method for Studying the Fold Surface of Space Targets
Weiwei Sun, and Lei Ding
To obtain high quality simulation images of a space target, simulating the folded surface of the target covered temperature-controlled material is important. This study employed the inverse ray tracing method to establish a radiation transfer model. The aim of this study is to calculate the secondary reflection that ocTo obtain high quality simulation images of a space target, simulating the folded surface of the target covered temperature-controlled material is important. This study employed the inverse ray tracing method to establish a radiation transfer model. The aim of this study is to calculate the secondary reflection that occurs on the folded surface of a polyimide material and the Fresnel reflection of the material. The cosine weight importance sampling method was used to optimize the ray interval during the calculation of the secondary reflection. Consequently, the bidirectional reflection distribution function (BRDF) model, which can describe the Fresnel reflection, was improved. Using a three-dimensional software program for generating and quantifying the fold plane, the root mean square error of importance sampling on severe folds was reduced by 63.42% compared with uniform sampling. Moreover, compared with the measured BRDF model, the improved BRDF model can better describe the Fresnel reflection and discrete strong mirror reflection spots were observed in the simulation image of the fold plane. Furthermore, the simulated satellite image shows the detailed texture of the folded surface effectively, and the shadow effect is realistic, which can provide image for the research of space target visual navigation algorithm..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2428004 (2023)
Dual-Stream Feature Aggregation Network for Unmanned Aerial Vehicle Aerial Images Semantic Segmentation
Runzeng Li, Zaifeng Shi, Fanning Kong, Xiangyang Zhao, and Tao Luo
Large object size difference in unmanned aerial vehicle (UAV) aerial photography makes it difficult to take into account the segmentation effect of objects of different sizes in the receptive field. A dual-stream feature aggregation network (DSFA-Net) with two branches to extract low-level and high-level features separLarge object size difference in unmanned aerial vehicle (UAV) aerial photography makes it difficult to take into account the segmentation effect of objects of different sizes in the receptive field. A dual-stream feature aggregation network (DSFA-Net) with two branches to extract low-level and high-level features separately, is proposed for such problems. In the encoder, a low-level information extraction branch with three serial ConvNeXt modules is used to preserve more low-level features by generating more channels of features. In the deep feature branch, the coordinate attention atrous spatial pyramid pooling (CA-ASPP) module reassigns weights to feature maps in the channel dimension. It makes the module focus on segmentation objects of different sizes and deep-level multi-scale features are obtained. During the decoding process, the bilateral guided aggregation module performs resolution aggregation between the low-level and deep-level features. Our method is evaluated on the AeroScapes and Semantic Drone datasets, the mean intersection over union is 83.16% and 72.09% respectively, and the mean pixel accuracy is 90.75% and 80.34% respectively. The proposed method is more capable of segmenting objects with large difference sizes compared to mainstream methods. It is suitable for semantic segmentation tasks for UAV aerial images..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2428005 (2023)
Reviews
Research Status and Prospect of Night Image Dehazing Algorithm
Xia Liu, and Changlun Hou
Night image dehazing technology has become an important research content in the field of image processing technology. It has important significance for target tracking detection, video surveillance, remote sensing and so on. Haze images at night usually have the characteristics of low contrast, uneven illumination, colNight image dehazing technology has become an important research content in the field of image processing technology. It has important significance for target tracking detection, video surveillance, remote sensing and so on. Haze images at night usually have the characteristics of low contrast, uneven illumination, color offset, etc., which makes haze removal for night images face great challenges. Through summarizing the research status of night image de-fogging algorithms at home and abroad in recent years, the classical algorithms from the perspective of the physical model, non-physical model and deep learning are summarized, and the algorithm process, advantages and disadvantages are elaborated. Finally, the future research direction of the night fog removal algorithm is prospected..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2400002 (2023)
Application of Terahertz Technology in Cultural Heritage Conversation
Chenyu Li, and Liang Qu
Terahertz technology is widely used in the field of nondestructive testing owing to its unique characteristics such as excellent perspective, low energy, high spectral resolution, and exceptional time resolution. The terahertz technology is particularly useful in obtaining the spectral information of precious cultural Terahertz technology is widely used in the field of nondestructive testing owing to its unique characteristics such as excellent perspective, low energy, high spectral resolution, and exceptional time resolution. The terahertz technology is particularly useful in obtaining the spectral information of precious cultural relics and can confirm the defect positions in their internal structure. These results can provide the technical support for subsequent conservation of cultural relics. This paper summarizes the domestic and international application prospects of terahertz spectroscopy and imaging technology in nondestructive and in situ testing. Terahertz technology can satisfy the requirements of national conversation, which is contributed to solve the exact problem of domestic cultural relic restoration. It has a certain promotion value and is of considerable significance for national conversation..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2400003 (2023)
Research Progress on Laser-Induced Breakdown Spectroscopy in Biomedicine
Zehai Hou, Lianbo Guo, Weiliang Wang, Yanwu Chu, and Furi Lin
Minor changes to elements in organisms can have a direct impact on their metabolism and physiological processes. Rapid and accurate qualitative and quantitative analyses of these elements are critical for both metabolism detection and clinical disease diagnosis. As medical detection technology continues to develop and Minor changes to elements in organisms can have a direct impact on their metabolism and physiological processes. Rapid and accurate qualitative and quantitative analyses of these elements are critical for both metabolism detection and clinical disease diagnosis. As medical detection technology continues to develop and clinical demand continues to increase, researchers are seeking newer, faster, and more adaptable clinical analysis and diagnostic technologies. Because of its fast, real-time, and multi-element simultaneous detection capabilities, laser-induced breakdown spectroscopy (LIBS) technology has shown great promise in recent years for use in blood and pathological tissue detection and elemental distribution imaging. This paper provides a comprehensive review of the current research status and latest progress of LIBS technology in biomedical applications and evaluates the challenges and opportunities of LIBS technology in the application fields of blood detection, biological tissue analysis, and elemental imaging. The paper also provides suggestions for further promoting the application of LIBS technology in the biomedical field..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2400004 (2023)
Current Developments in Photoacoustic Imaging Technologies for Cultural Heritage Conservation
Yaowen Zhu, Xingyu Lin, and Yingjie Yu
Photoacoustic imaging technology combines the high contrast and high resolution properties of optics and the high penetration depth property of acoustics. It has developed into a distinctive and advantageous non-destructive inspection and imaging technology that uses ultrasonic signals generated by the photoacoustic efPhotoacoustic imaging technology combines the high contrast and high resolution properties of optics and the high penetration depth property of acoustics. It has developed into a distinctive and advantageous non-destructive inspection and imaging technology that uses ultrasonic signals generated by the photoacoustic effect to resolve depth information, thus locating objects in a three-dimensional space. In the field of cultural heritage conservation, materials used to make cultural artifacts have different optical absorption properties; thus, ultrasonic signals with different amplitudes and frequencies are generated and radiated outward when they are subjected to laser light excitation. The characteristics of ultrasonic signals can reveal the material type and detect surface defects; hence, photoacoustic imaging technology can be applied in cultural heritage conservation. This paper provides an overview of existing photoacoustic imaging technologies, describes their current developments and limitations for cultural heritage conservation, and summarizes the future prospects of the technology for these applications..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2400005 (2023)
Spectroscopy
Influencing Factors of Mid-Infrared Spectrum Blood Glucose Detection
Yansong Yue, Zhushanying Zhang, Sicong Zhu, Huimin Cao, Dongyun Zheng, and Qinlan Xie
Mid-infrared attenuated total reflection spectroscopy has the natural advantage of fast and green blood glucose detection in humans. However, the presence of other components in human blood can affect the accuracy of glucose detection. Therefore, we study the interference degree of the presence of cholesterol, albumin Mid-infrared attenuated total reflection spectroscopy has the natural advantage of fast and green blood glucose detection in humans. However, the presence of other components in human blood can affect the accuracy of glucose detection. Therefore, we study the interference degree of the presence of cholesterol, albumin and urea in human blood on the detection of blood glucose by infrared spectroscopy. Taking 117 parts of glucose mimicry solution containing different interferences and different mass concentrations as the research object, the original spectrum is smoothed by Savitzky-Golay to construct a partial least squares regression model, and the Clarke Error Grid and comparison plot of predicted value and true value are constructed for further analysis. The results show that the prediction set correlation coefficient (Rp) and root mean square error (RMSEP) of the prediction set of the total interferer model are 0.9785 and 40.0187, respectively, and 85.7% of the prediction results fall in the Clarke Error Grid A region. The Rp and RMSEP of the missing cholesterol model are 0.9042 and 175.7292, respectively, and 40% of the predictions fall in region A. The Rp and RMSEP of the missing albumin model are 0.9616 and 103.6627, respectively, and 42.9% of the predictions fall in region A. The Rp and RMSEP of the urea deletion model are 0.9742 and 38.6716, respectively, and all predictions fall in region A. It can be seen that cholesterol has the greatest degree of interference, followed by albumin, and urea produces the least interference. This study has certain help and reference value for improving the accuracy of glucose detection by infrared spectroscopy..
Laser & Optoelectronics Progress
- Publication Date: Dec. 25, 2023
- Vol. 60, Issue 24, 2430001 (2023)
© Copyright 2018-2021 | Chinese Laser Press. All Rights Reserved 沪ICP备15018463号-20