Pub Date : 2022-11-01DOI: 10.1109/CCISP55629.2022.9974245
Mu Qiao, Ying Lin, Meng Liu, Zhuangzhuang Li, Wenjie Zheng, Yi Yang, Xu Jiang
The routine inspection of an electricity substation helps to detect faults and repair equipment in time, ensuring the substation to work safely. However, due to irregular operations, some inspectors may miss to capture images at certain spots while take multiple similar images at the same spots. In order to make control of the inspection quality, we design an algorithm to find such situation automatically. Specifically, given two images, we design a registration-based method to evaluate the affine transformation between two images, then we evaluate the averaged corner error by comparing the image transformed with respect to the estimated affine transformation to an identical transformation. Finally, we screen out the similar images that are small in the averaged corner error. These images are very likely to be taken at the same inspection spot. We conduct experiments on a dataset collected during one routine inspection of a whole substation. Experimental results show that our method is effective to screen out similar images, helping to build an automatic quality control process of the routine inspection.
{"title":"Image Similarity Measurement for The Quality Control of Electricity Substation Inspection","authors":"Mu Qiao, Ying Lin, Meng Liu, Zhuangzhuang Li, Wenjie Zheng, Yi Yang, Xu Jiang","doi":"10.1109/CCISP55629.2022.9974245","DOIUrl":"https://doi.org/10.1109/CCISP55629.2022.9974245","url":null,"abstract":"The routine inspection of an electricity substation helps to detect faults and repair equipment in time, ensuring the substation to work safely. However, due to irregular operations, some inspectors may miss to capture images at certain spots while take multiple similar images at the same spots. In order to make control of the inspection quality, we design an algorithm to find such situation automatically. Specifically, given two images, we design a registration-based method to evaluate the affine transformation between two images, then we evaluate the averaged corner error by comparing the image transformed with respect to the estimated affine transformation to an identical transformation. Finally, we screen out the similar images that are small in the averaged corner error. These images are very likely to be taken at the same inspection spot. We conduct experiments on a dataset collected during one routine inspection of a whole substation. Experimental results show that our method is effective to screen out similar images, helping to build an automatic quality control process of the routine inspection.","PeriodicalId":431851,"journal":{"name":"2022 7th International Conference on Communication, Image and Signal Processing (CCISP)","volume":"72 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115812348","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-01DOI: 10.1109/CCISP55629.2022.9974203
Wei Li, Dandan Li, Jiahao Jiang, Yang Gao
Millimeter wave radar has the advantage of all-weather and all-day detection in a variety of scenarios, but its low angular resolution makes it difficult to meet the demand for three-dimensional imaging. In this paper, a 64 GHz MIMO millimeter wave radar with 20 Tx antennas and 20 Rx antennas is used to improve the angular resolution. The radar raw signal is first processed by a three-dimensional fast Fourier Transform (3D-FFT) algorithm. Then the target is extracted using the dual-channel constant false alarm detection (CFAR) algorithm, which can discriminate the cloud points in two dimensions simultaneously. Through bilinear interpolation of the cloud of points, a 3D contour image can be obtained. The radar imaging experiment is carried out with the metal ball pile as the target. The experimental results show that the radar can obtain a high resolution three-dimensional contour image.
{"title":"3D contour imaging based on a Millimeter wave MIMO radar","authors":"Wei Li, Dandan Li, Jiahao Jiang, Yang Gao","doi":"10.1109/CCISP55629.2022.9974203","DOIUrl":"https://doi.org/10.1109/CCISP55629.2022.9974203","url":null,"abstract":"Millimeter wave radar has the advantage of all-weather and all-day detection in a variety of scenarios, but its low angular resolution makes it difficult to meet the demand for three-dimensional imaging. In this paper, a 64 GHz MIMO millimeter wave radar with 20 Tx antennas and 20 Rx antennas is used to improve the angular resolution. The radar raw signal is first processed by a three-dimensional fast Fourier Transform (3D-FFT) algorithm. Then the target is extracted using the dual-channel constant false alarm detection (CFAR) algorithm, which can discriminate the cloud points in two dimensions simultaneously. Through bilinear interpolation of the cloud of points, a 3D contour image can be obtained. The radar imaging experiment is carried out with the metal ball pile as the target. The experimental results show that the radar can obtain a high resolution three-dimensional contour image.","PeriodicalId":431851,"journal":{"name":"2022 7th International Conference on Communication, Image and Signal Processing (CCISP)","volume":"131 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124128473","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-01DOI: 10.1109/CCISP55629.2022.9974473
Bo Wen, Tzu-kao Wang
Passive terahertz radiometer scanner [1] is an emerging type of handheld security inspection device that could overcome some of the shortcomings of current security inspection devices on the market. However, subject to several difficulties such as unstable measurements and ambiguous signal features, to detect hidden objects using this device is challenging. The previous research on this topic was insufficient, and the object detection algorithm was less reliable and lacked scientific verifi-cation. In this paper, we propose a whole new pipeline to address this task. We explore and compare a series of adaptive filtering techniques and propose a customized Kalman filter to extract the signal features that describe hidden objects. Then, we adopt two machine learning methods on the filtered signal to detect the hidden objects. Experiment shows that the proposed pipeline can achieve over 85 % accuracy, which hugely outperforms the old methods.
{"title":"Passive THz Radiometer Scanner Object Detection with Adaptive Filtering","authors":"Bo Wen, Tzu-kao Wang","doi":"10.1109/CCISP55629.2022.9974473","DOIUrl":"https://doi.org/10.1109/CCISP55629.2022.9974473","url":null,"abstract":"Passive terahertz radiometer scanner [1] is an emerging type of handheld security inspection device that could overcome some of the shortcomings of current security inspection devices on the market. However, subject to several difficulties such as unstable measurements and ambiguous signal features, to detect hidden objects using this device is challenging. The previous research on this topic was insufficient, and the object detection algorithm was less reliable and lacked scientific verifi-cation. In this paper, we propose a whole new pipeline to address this task. We explore and compare a series of adaptive filtering techniques and propose a customized Kalman filter to extract the signal features that describe hidden objects. Then, we adopt two machine learning methods on the filtered signal to detect the hidden objects. Experiment shows that the proposed pipeline can achieve over 85 % accuracy, which hugely outperforms the old methods.","PeriodicalId":431851,"journal":{"name":"2022 7th International Conference on Communication, Image and Signal Processing (CCISP)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126422159","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The rapid social development has given rise to a growing concern over environmental issues, one of which is the disposal of solid waste. Recycling is considered as one of the critical disposal methods. Taking into consideration of fast, intelligent classification and identification of the solid waste as a prerequisite for recycling and utilization, a multiple material feature based solid waste identification and classification method is proposed in this paper. The experimental results show that the proposed method achieves an accuracy of 83.7% on an in-house textile solid waste image dataset. The results indicates that our method with multiple material features is able to handle the textile solid waste recognition problem properly.
{"title":"Textile Solid Waste Recognition with Multiple Material Features","authors":"Yuan Gou, Wei Dong, Lin Gan, Ling He, Wanyu Tang, Jing Zhang","doi":"10.1109/CCISP55629.2022.9974371","DOIUrl":"https://doi.org/10.1109/CCISP55629.2022.9974371","url":null,"abstract":"The rapid social development has given rise to a growing concern over environmental issues, one of which is the disposal of solid waste. Recycling is considered as one of the critical disposal methods. Taking into consideration of fast, intelligent classification and identification of the solid waste as a prerequisite for recycling and utilization, a multiple material feature based solid waste identification and classification method is proposed in this paper. The experimental results show that the proposed method achieves an accuracy of 83.7% on an in-house textile solid waste image dataset. The results indicates that our method with multiple material features is able to handle the textile solid waste recognition problem properly.","PeriodicalId":431851,"journal":{"name":"2022 7th International Conference on Communication, Image and Signal Processing (CCISP)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126846473","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-01DOI: 10.1109/CCISP55629.2022.9974449
Ma Hongbao, Kang Yihua, Cai Xiang, Qiu Gongzhe, Cheng Si, Jin Xin
Ultrasonic Non-Destructive Evaluation (NDE) has been proven to be an effective means to assure the measurement of material properties. However, accurate detection of defect echoes buried in strong noise is challenging. A novel de-noising method based on S-transform and Non-negative matrix factorization is proposed in this paper. In the first stage, the S-transform was performed on the original signal to obtain the time-frequency distribution. Subsequently, the feature separation of echo signal and noise is realized by non-negative matrix decomposition. Finally, clear denoising defect waveforms are acquired by the inverse S-transform. Both simulation analysis and experimental results show the effectiveness and superiority of the proposed method in noise suppression of ultrasonic NDE.
{"title":"Denoising Ultrasonic Echo Signals with S-Transform and Non-negative matrix factorization","authors":"Ma Hongbao, Kang Yihua, Cai Xiang, Qiu Gongzhe, Cheng Si, Jin Xin","doi":"10.1109/CCISP55629.2022.9974449","DOIUrl":"https://doi.org/10.1109/CCISP55629.2022.9974449","url":null,"abstract":"Ultrasonic Non-Destructive Evaluation (NDE) has been proven to be an effective means to assure the measurement of material properties. However, accurate detection of defect echoes buried in strong noise is challenging. A novel de-noising method based on S-transform and Non-negative matrix factorization is proposed in this paper. In the first stage, the S-transform was performed on the original signal to obtain the time-frequency distribution. Subsequently, the feature separation of echo signal and noise is realized by non-negative matrix decomposition. Finally, clear denoising defect waveforms are acquired by the inverse S-transform. Both simulation analysis and experimental results show the effectiveness and superiority of the proposed method in noise suppression of ultrasonic NDE.","PeriodicalId":431851,"journal":{"name":"2022 7th International Conference on Communication, Image and Signal Processing (CCISP)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130837930","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-01DOI: 10.1109/CCISP55629.2022.9974244
Kaijun Wu, Bo Tian, Yougang Wen, Xue Wang
The convolution neural network (CNN) is vulnerable to the adversarial attack, because the attack can generate adversarial images to force the CNN to misclassify the original label of the clean image. To defend against the adversarial attack, we propose to detect the adversarial images first and then prevent feeding the adversarial image into the CNN model. In this paper, we employ a steganalysis based method based on rich residual models to detect adversarial images which are generated by the typical attacks including BIM and DEEPFOOL. The rich residual models not only reduce the influences from natural image contents, but also enhance the diversity of the feature. Two typical and complementary methods spatial rich model (SRM) and projected spatial rich model (PSRM) are used to extract the feature, where SRM finely capture the statistical changes on co-occurrence in a small neighborhood, and PSRM remedy the loss information caused by SRM. Experimental results on CIFAR-IO and ImageNet show that the proposed method obtained better performance than existing steganalysis methods for detecting adversarial images generated by BIM and DEEPFOOL attack. The research results are expected to improve the recognition ability of image adversarial samples in the convolutional neural network model, and support the data security of natural image content in image recognition.
{"title":"Detecting Adversarial Examples Using Rich Residual Models to Improve Data Security in CNN Models","authors":"Kaijun Wu, Bo Tian, Yougang Wen, Xue Wang","doi":"10.1109/CCISP55629.2022.9974244","DOIUrl":"https://doi.org/10.1109/CCISP55629.2022.9974244","url":null,"abstract":"The convolution neural network (CNN) is vulnerable to the adversarial attack, because the attack can generate adversarial images to force the CNN to misclassify the original label of the clean image. To defend against the adversarial attack, we propose to detect the adversarial images first and then prevent feeding the adversarial image into the CNN model. In this paper, we employ a steganalysis based method based on rich residual models to detect adversarial images which are generated by the typical attacks including BIM and DEEPFOOL. The rich residual models not only reduce the influences from natural image contents, but also enhance the diversity of the feature. Two typical and complementary methods spatial rich model (SRM) and projected spatial rich model (PSRM) are used to extract the feature, where SRM finely capture the statistical changes on co-occurrence in a small neighborhood, and PSRM remedy the loss information caused by SRM. Experimental results on CIFAR-IO and ImageNet show that the proposed method obtained better performance than existing steganalysis methods for detecting adversarial images generated by BIM and DEEPFOOL attack. The research results are expected to improve the recognition ability of image adversarial samples in the convolutional neural network model, and support the data security of natural image content in image recognition.","PeriodicalId":431851,"journal":{"name":"2022 7th International Conference on Communication, Image and Signal Processing (CCISP)","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125597244","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-01DOI: 10.1109/CCISP55629.2022.9974352
Guilan Luo, Anqi Cao, Shuailin Wang, Yuan-Xin Zhu
In order to solve the problem of low node allocation efficiency of UAV swarms in forest fire disaster relief scenarios, a weighted multi-objective particle swarm UAV swarm scheduling algorithm was proposed, and the visualization of simulation scheduling was realized. Through the improvement of the target point allocation model, the standardization of the target point weight, and the definition of the comprehensive evaluation index of the UAV, after selecting the UAV with the optimal performance and the largest distribution probability, the remaining UAVs are distributed according to the average probability. Scheduling to improve the real-time performance of UAV swarm scheduling. Finally, through the simulation experiment and performance analysis of the simulation system, the results show that the improved algorithm of UAV swarm scheduling average convergence time is reduced by about 30s compared with the original algorithm, has better convergence, and the UAV swarm scheduling efficiency is improved.
{"title":"UA V Swarm Scheduling Based on Weighted Multi-Objective Particle Swarm Algorithm","authors":"Guilan Luo, Anqi Cao, Shuailin Wang, Yuan-Xin Zhu","doi":"10.1109/CCISP55629.2022.9974352","DOIUrl":"https://doi.org/10.1109/CCISP55629.2022.9974352","url":null,"abstract":"In order to solve the problem of low node allocation efficiency of UAV swarms in forest fire disaster relief scenarios, a weighted multi-objective particle swarm UAV swarm scheduling algorithm was proposed, and the visualization of simulation scheduling was realized. Through the improvement of the target point allocation model, the standardization of the target point weight, and the definition of the comprehensive evaluation index of the UAV, after selecting the UAV with the optimal performance and the largest distribution probability, the remaining UAVs are distributed according to the average probability. Scheduling to improve the real-time performance of UAV swarm scheduling. Finally, through the simulation experiment and performance analysis of the simulation system, the results show that the improved algorithm of UAV swarm scheduling average convergence time is reduced by about 30s compared with the original algorithm, has better convergence, and the UAV swarm scheduling efficiency is improved.","PeriodicalId":431851,"journal":{"name":"2022 7th International Conference on Communication, Image and Signal Processing (CCISP)","volume":"31 10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134488026","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-01DOI: 10.1109/CCISP55629.2022.9974237
Shiying Yao, Xuanyu Yang, Z. Song, Xiaolong Yang, Dengwei Duan, Haizhou Yang
To ensure the security and reliable transmission of private information in Internet of Things for Smart Grid (for short SG-IoT), we proposed a privacy-aware secure routing method (i.e., Maze Routing) based on a novel directional random routing paradigm, which consists of a direction-determined virtual global routing mode and a physical local routing mode. Firstly, for an in-transit packet at the current node, all k-hop nodes are defined as its k-hop wavefronts, and many x-hop wavefronts within k hops are further gathered as a hyper next-hop set. Then, under the direction-determined virtual global routing mode, the packet randomly selects one of elements from the next-hop set by a priority probability as its global next-hop node according to its privacy awareness and End-to-End Quality of Service (E2E QoS) requirement, so as to build a virtual global routing path with a definite direction determined by the destination of this packet. Moreover, under the physical local routing mode, a local pathlet for any two consecutive global next-hop nodes can be built in an ordinary shortest-first routing scheme, then such many pathlets are wired in sequence to constitute a path in whole from source node to destination. Under several loT application scenarios within smart grid, the simulations results show that the proposed maze routing outperforms the existing shortest-first routing method in terms of privacy protection, E2E QoS and communication overhead.
为了保证智能电网物联网(简称SG-IoT)中私有信息的安全可靠传输，我们提出了一种基于新型定向随机路由范式的隐私感知安全路由方法(即迷宫路由)，该方法由方向确定的虚拟全局路由模式和物理本地路由模式组成。首先，对于当前节点的传输数据包，将所有k-hop节点定义为其k-hop波前，并将k跳内的许多x-hop波前进一步收集为超下一跳集。然后，在方向确定的虚拟全局路由模式下，数据包根据其隐私意识和端到端服务质量(End-to-End Quality of Service, E2E QoS)要求，从按优先级概率设置的下一跳元素中随机选择一个元素作为其全局下一跳节点，从而构建由数据包目的地决定方向明确的虚拟全局路由路径。此外，在物理本地路由模式下，普通的最短优先路由方案可以为任意两个连续的全局下一跳节点建立一个本地路径，然后将这些路径按顺序连接，构成从源节点到目的节点的一条完整路径。在智能电网的多个loT应用场景下，仿真结果表明，本文提出的迷宫路由在隐私保护、端到端QoS和通信开销方面优于现有的最短优先路由方法。
{"title":"Maze Routing: An Information Privacy-aware Secure Routing in Internet of Things for Smart Grid","authors":"Shiying Yao, Xuanyu Yang, Z. Song, Xiaolong Yang, Dengwei Duan, Haizhou Yang","doi":"10.1109/CCISP55629.2022.9974237","DOIUrl":"https://doi.org/10.1109/CCISP55629.2022.9974237","url":null,"abstract":"To ensure the security and reliable transmission of private information in Internet of Things for Smart Grid (for short SG-IoT), we proposed a privacy-aware secure routing method (i.e., Maze Routing) based on a novel directional random routing paradigm, which consists of a direction-determined virtual global routing mode and a physical local routing mode. Firstly, for an in-transit packet at the current node, all k-hop nodes are defined as its k-hop wavefronts, and many x-hop wavefronts within k hops are further gathered as a hyper next-hop set. Then, under the direction-determined virtual global routing mode, the packet randomly selects one of elements from the next-hop set by a priority probability as its global next-hop node according to its privacy awareness and End-to-End Quality of Service (E2E QoS) requirement, so as to build a virtual global routing path with a definite direction determined by the destination of this packet. Moreover, under the physical local routing mode, a local pathlet for any two consecutive global next-hop nodes can be built in an ordinary shortest-first routing scheme, then such many pathlets are wired in sequence to constitute a path in whole from source node to destination. Under several loT application scenarios within smart grid, the simulations results show that the proposed maze routing outperforms the existing shortest-first routing method in terms of privacy protection, E2E QoS and communication overhead.","PeriodicalId":431851,"journal":{"name":"2022 7th International Conference on Communication, Image and Signal Processing (CCISP)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134570369","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-01DOI: 10.1109/CCISP55629.2022.9974164
Yang Li, Sanxin Jiang
To realize the automatic detection of wafer surface defects, we propose Skip-MemGANs unsupervised detection network, which is an ensemble generative adversarial network that automatically detects defects by the difference between the target image and the reconstructed image.The network is composed of three generators and three discriminators. Each generator uses encoder-decoder convolutional neural network with two layers of skip connection and memory module to capture multi-scale input image features. These generators are randomly paired with discriminators, and receive feedback from the three discriminators, while the discriminators receive reconstructed samples from the three generators.Compared with a single GAN, the ensemble GAN can better simulate the distribution of normal data in the high-dimensional image space.We evaluate the single GAN model, GAN ensemble model and other basic models. The results show that our proposed Skip-MemGANs network outperforms other models in wafer defect detection task, the AUC value reached 0.956.
{"title":"Skip-MemGANs: An Ensemble Generative Adversarial Network Based on Skip Connection and Memory Module for Wafer Defect Detection","authors":"Yang Li, Sanxin Jiang","doi":"10.1109/CCISP55629.2022.9974164","DOIUrl":"https://doi.org/10.1109/CCISP55629.2022.9974164","url":null,"abstract":"To realize the automatic detection of wafer surface defects, we propose Skip-MemGANs unsupervised detection network, which is an ensemble generative adversarial network that automatically detects defects by the difference between the target image and the reconstructed image.The network is composed of three generators and three discriminators. Each generator uses encoder-decoder convolutional neural network with two layers of skip connection and memory module to capture multi-scale input image features. These generators are randomly paired with discriminators, and receive feedback from the three discriminators, while the discriminators receive reconstructed samples from the three generators.Compared with a single GAN, the ensemble GAN can better simulate the distribution of normal data in the high-dimensional image space.We evaluate the single GAN model, GAN ensemble model and other basic models. The results show that our proposed Skip-MemGANs network outperforms other models in wafer defect detection task, the AUC value reached 0.956.","PeriodicalId":431851,"journal":{"name":"2022 7th International Conference on Communication, Image and Signal Processing (CCISP)","volume":"165 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115295919","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-01DOI: 10.1109/CCISP55629.2022.9974287
Yudong Wang, Zhou Yi
Reconfigurable optical splitter multiplexer (ROADM) devices are the core devices for current optical network interconnection and dense wavelength division multiplexing. In this paper, the Google Remote Procedure Call (gRPC) protocol is investigated for the shortcomings of low transmission rate, poor transmission efficiency and only local calls when communicating with RESTful and TCP protocols mainly used in the current ROADM device software calling system. A reconfigurable remote procedure call framework for optical splitter and multiplexer based on the gRPC protocol was implemented using C language. The shortcomings of the ROADM device call system in terms of transfer rate, transfer efficiency and the fact that only local procedure calls can be made have been improved. Test results show that the use of the gRPC protocol has effectively improved the transmission rate and efficiency of the RODAM device and enabled remote procedure calls.
{"title":"Reconfigurable optical demultiplexer calling framework based on gRPC Design","authors":"Yudong Wang, Zhou Yi","doi":"10.1109/CCISP55629.2022.9974287","DOIUrl":"https://doi.org/10.1109/CCISP55629.2022.9974287","url":null,"abstract":"Reconfigurable optical splitter multiplexer (ROADM) devices are the core devices for current optical network interconnection and dense wavelength division multiplexing. In this paper, the Google Remote Procedure Call (gRPC) protocol is investigated for the shortcomings of low transmission rate, poor transmission efficiency and only local calls when communicating with RESTful and TCP protocols mainly used in the current ROADM device software calling system. A reconfigurable remote procedure call framework for optical splitter and multiplexer based on the gRPC protocol was implemented using C language. The shortcomings of the ROADM device call system in terms of transfer rate, transfer efficiency and the fact that only local procedure calls can be made have been improved. Test results show that the use of the gRPC protocol has effectively improved the transmission rate and efficiency of the RODAM device and enabled remote procedure calls.","PeriodicalId":431851,"journal":{"name":"2022 7th International Conference on Communication, Image and Signal Processing (CCISP)","volume":"532 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123935984","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：