Pub Date : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874224
Xiaoqi Luo, Yuanjie Xing, Senhai Xu
Super pixel algorithm SLIC uses K-means mean clustering method to effectively generate super pixels. Compared with other super pixel algorithms, it is more efficient and improves the segmentation performance. In order to further improve its performance, the program is optimized from five major directions: compilation optimization, data structure optimization, loop vectorization, OpenMP parallel optimization and algorithm optimization.
{"title":"Parallel Optimization of Super Pixel Algorithm SLIC","authors":"Xiaoqi Luo, Yuanjie Xing, Senhai Xu","doi":"10.1109/ISPDS56360.2022.9874224","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874224","url":null,"abstract":"Super pixel algorithm SLIC uses K-means mean clustering method to effectively generate super pixels. Compared with other super pixel algorithms, it is more efficient and improves the segmentation performance. In order to further improve its performance, the program is optimized from five major directions: compilation optimization, data structure optimization, loop vectorization, OpenMP parallel optimization and algorithm optimization.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"8 11","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120866540","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874052
Lu Chen, Yang Bai, Q. Cheng, Mei Wu
Despite the many advantages of Convolutional Neural Networks (CNN), their perceptual fields are usually small and not conducive to capturing global features. In contrast, Transformer is able to capture long-range dependencies and obtain global information of an image with self-attention. For combining the advantages of CNN and Transformer, we propose to integrate the Local Aggregation module to the structure of Swin Transformer. The Local Aggregation module includes lightweight Depthwise Convolution and Pointwise Convolution, and it can locally capture the information of feature map at stages of Swin Transformer. Our experiments demonstrate that accuracy can be improved with such an integrated model. On the Cifar-10 dataset, the Top-1 accuracy reaches 87.74%, which is 3.32% higher than Swin, and the Top-5 accuracy reaches 99.54%; on the Mini-ImageNet dataset, the Top-1 accuracy reaches 79.1%, which is 7.68% higher than Swin, and the Top-5 accuracy reaches 94.02%, which is 3.25% higher than Swin 3.25%.
{"title":"Swin Transformer with Local Aggregation","authors":"Lu Chen, Yang Bai, Q. Cheng, Mei Wu","doi":"10.1109/ISPDS56360.2022.9874052","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874052","url":null,"abstract":"Despite the many advantages of Convolutional Neural Networks (CNN), their perceptual fields are usually small and not conducive to capturing global features. In contrast, Transformer is able to capture long-range dependencies and obtain global information of an image with self-attention. For combining the advantages of CNN and Transformer, we propose to integrate the Local Aggregation module to the structure of Swin Transformer. The Local Aggregation module includes lightweight Depthwise Convolution and Pointwise Convolution, and it can locally capture the information of feature map at stages of Swin Transformer. Our experiments demonstrate that accuracy can be improved with such an integrated model. On the Cifar-10 dataset, the Top-1 accuracy reaches 87.74%, which is 3.32% higher than Swin, and the Top-5 accuracy reaches 99.54%; on the Mini-ImageNet dataset, the Top-1 accuracy reaches 79.1%, which is 7.68% higher than Swin, and the Top-5 accuracy reaches 94.02%, which is 3.25% higher than Swin 3.25%.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"61 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124948914","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874034
Shengchen Wang, Xisheng Li, Wenyu Huo, Jia You
In order to improve the fusion quality of infrared and visible light images, enhance the visual effect of fused images, and solve the problems that traditional fusion methods need to manually set fusion rules and the background details of fused images are poorly preserved, this paper proposes an improved generative adversarial network that combines multi-scale information. The generator used in this method is a typical encoder and decoder structure, and the discriminator uses a dual discriminator to establish the confrontation relationship between the infrared source image, the visible light source image and the fusion image respectively. Before the source image is input to the encoder, multi-scale information is introduced through the Inception network, which effectively extracts the multi-scale features of the image, which ensures the subsequent improvement of the quality of the fusion image. In addition, the loss function is improved to retain more background details and highlight infrared feature information. The control experiment results show that the method in this paper obtains better fusion effect in subjective and objective evaluation.
{"title":"Fusion of Infrared and Visible Images Based on Improved Generative Adversarial Networks","authors":"Shengchen Wang, Xisheng Li, Wenyu Huo, Jia You","doi":"10.1109/ISPDS56360.2022.9874034","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874034","url":null,"abstract":"In order to improve the fusion quality of infrared and visible light images, enhance the visual effect of fused images, and solve the problems that traditional fusion methods need to manually set fusion rules and the background details of fused images are poorly preserved, this paper proposes an improved generative adversarial network that combines multi-scale information. The generator used in this method is a typical encoder and decoder structure, and the discriminator uses a dual discriminator to establish the confrontation relationship between the infrared source image, the visible light source image and the fusion image respectively. Before the source image is input to the encoder, multi-scale information is introduced through the Inception network, which effectively extracts the multi-scale features of the image, which ensures the subsequent improvement of the quality of the fusion image. In addition, the loss function is improved to retain more background details and highlight infrared feature information. The control experiment results show that the method in this paper obtains better fusion effect in subjective and objective evaluation.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"175 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125800400","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874216
Qingtong Yang, Yuanxun Fan
In view of the shortcomings of pipeline cleaning robot in the domestic market. A new pipeline cleaning robot is designed. The design idea is to combine high-pressure water jet cleaning equipment with the main body of the robot, equipped with underwater camera and image processing system. A cleaning effect evaluation method based on image processing, using python language is designed and applied to the robot. After a series of image pre-processing, including defogging, image enhancement, image segmentation, image binarization, etc. of the collected images, black-and-white images are obtained which clearly distinguish the scale and the pipe wall. The cleaning effect is evaluated according to the proportion of black pixels and white pixels. The test results show that this method can effectively process the original fogged image to distinguish the scale and pipe wall in the image. This method provides a basis for the evaluation of the operation effect of the pipe cleaning robot.
{"title":"An evaluation method of municipal pipeline cleaning effect based on image processing","authors":"Qingtong Yang, Yuanxun Fan","doi":"10.1109/ISPDS56360.2022.9874216","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874216","url":null,"abstract":"In view of the shortcomings of pipeline cleaning robot in the domestic market. A new pipeline cleaning robot is designed. The design idea is to combine high-pressure water jet cleaning equipment with the main body of the robot, equipped with underwater camera and image processing system. A cleaning effect evaluation method based on image processing, using python language is designed and applied to the robot. After a series of image pre-processing, including defogging, image enhancement, image segmentation, image binarization, etc. of the collected images, black-and-white images are obtained which clearly distinguish the scale and the pipe wall. The cleaning effect is evaluated according to the proportion of black pixels and white pixels. The test results show that this method can effectively process the original fogged image to distinguish the scale and pipe wall in the image. This method provides a basis for the evaluation of the operation effect of the pipe cleaning robot.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114925646","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874083
Ning Zhou, Zhengxin Liu, Jianxin Zhou
Micropipe defects on the surface of silicon carbide wafers can have a significant impact on the quality of the wafers. Therefore, it is necessary to identify and locate them during the production process. Due to micropipe defects being small and dense, which are difficult to detect completely, we propose a real-time defect detection network model based on the Yolov5. The model adds a detection branch in the neck and head block of Yolov5 to obtain high-resolution features. To get the spatial and channel attention, we apply a CBAM attention module in each neck branch, and DA attention module in each head branch. The experiments show that our model improves AP by 1.89% and increases precision and recall by 10.12% and 2.95%, respectively, compared with the Yolov5 model. The results show that our model has a better ability to detect small and dense defects.
{"title":"Yolov5-based defect detection for wafer surface micropipe","authors":"Ning Zhou, Zhengxin Liu, Jianxin Zhou","doi":"10.1109/ISPDS56360.2022.9874083","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874083","url":null,"abstract":"Micropipe defects on the surface of silicon carbide wafers can have a significant impact on the quality of the wafers. Therefore, it is necessary to identify and locate them during the production process. Due to micropipe defects being small and dense, which are difficult to detect completely, we propose a real-time defect detection network model based on the Yolov5. The model adds a detection branch in the neck and head block of Yolov5 to obtain high-resolution features. To get the spatial and channel attention, we apply a CBAM attention module in each neck branch, and DA attention module in each head branch. The experiments show that our model improves AP by 1.89% and increases precision and recall by 10.12% and 2.95%, respectively, compared with the Yolov5 model. The results show that our model has a better ability to detect small and dense defects.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128687888","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874191
Guofang Zhang, Zihe Zhou, Hanlin Shao
At present, the main problems in the layout of shared electric vehicle sites are: many sites have problems in site selection, which has affected the profitability of operators, and some sites have poor site selection, which makes it difficult to make profits due to the small number of users. However, in some areas with high demands, sites have not been established yet, resulting in missed business opportunities. In order to deeply study the location method of shared electric vehicles, an innovative “AHP-MOP method” was created. This method uses the analytic hierarchy process to establish a set of site evaluation system, then carries out preliminary optimization based on the maximum coverage model, then carries out secondary optimization based on the multi-objective optimization model to obtain the final optimization results. The famous “car capital” Wuhan Zhuankou is selected as the main example location. Based on the above theoretical research, the rationality of the “AHP-MOP method” is further verified. Finally, 9 shared electric vehicle stations are successfully selected in Zhuankou. The results are scientific, reasonable and in line with the reality.
{"title":"Research on the layout of shared electric vehicle charging stations based on AHP-MOP method","authors":"Guofang Zhang, Zihe Zhou, Hanlin Shao","doi":"10.1109/ISPDS56360.2022.9874191","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874191","url":null,"abstract":"At present, the main problems in the layout of shared electric vehicle sites are: many sites have problems in site selection, which has affected the profitability of operators, and some sites have poor site selection, which makes it difficult to make profits due to the small number of users. However, in some areas with high demands, sites have not been established yet, resulting in missed business opportunities. In order to deeply study the location method of shared electric vehicles, an innovative “AHP-MOP method” was created. This method uses the analytic hierarchy process to establish a set of site evaluation system, then carries out preliminary optimization based on the maximum coverage model, then carries out secondary optimization based on the multi-objective optimization model to obtain the final optimization results. The famous “car capital” Wuhan Zhuankou is selected as the main example location. Based on the above theoretical research, the rationality of the “AHP-MOP method” is further verified. Finally, 9 shared electric vehicle stations are successfully selected in Zhuankou. The results are scientific, reasonable and in line with the reality.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"123 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126700179","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874012
Meizheng Ge, Qiong Liu
In engineering applications, traditional methods are usually used to acquire microscopic images, and due to objective factors such as uneven acquisition equipment and uneven illumination, there may be problems such as unclear acquisition images and insufficient local exposure. Traditional de-sharpening mask image enhancement algorithms are suitable for enhancing the edges and details of microscopic images, but are extremely sensitive to noise and do not enhance contrast and detail at the same time. In this paper, an improved sharpening masking algorithm is proposed, which sharpens the edges of rock images in the brightness channel of HSV color space, uses the difference between the non-local mean filter image and the original image to achieve high-frequency components, adaptively enhances the high-frequency components, improves the problem of image light unevenness through the contrast enhancement algorithm, and superimposes it with high-frequency components to effectively enhance the image. Experimental results show that the image processed by the algorithm has outstanding details and clear textures, which better suppresses the amplification of noise.
{"title":"Rock Image Contrast Enhancement Method Based on Improved De-sharpening Mask","authors":"Meizheng Ge, Qiong Liu","doi":"10.1109/ISPDS56360.2022.9874012","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874012","url":null,"abstract":"In engineering applications, traditional methods are usually used to acquire microscopic images, and due to objective factors such as uneven acquisition equipment and uneven illumination, there may be problems such as unclear acquisition images and insufficient local exposure. Traditional de-sharpening mask image enhancement algorithms are suitable for enhancing the edges and details of microscopic images, but are extremely sensitive to noise and do not enhance contrast and detail at the same time. In this paper, an improved sharpening masking algorithm is proposed, which sharpens the edges of rock images in the brightness channel of HSV color space, uses the difference between the non-local mean filter image and the original image to achieve high-frequency components, adaptively enhances the high-frequency components, improves the problem of image light unevenness through the contrast enhancement algorithm, and superimposes it with high-frequency components to effectively enhance the image. Experimental results show that the image processed by the algorithm has outstanding details and clear textures, which better suppresses the amplification of noise.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"138 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121588287","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874020
Pengfei Li, Heng Wang, Xueyu Huang
In agricultural production, the growth and yield of crops have always attracted people's attention. For the detection of wheat planting density, a wheat straw detection model based on improved YOLOv5 is proposed in this paper. Firstly, at the end of the backbone network, the C3 module (C3TR) integrated with Transformer is used to replace the traditional C3 module, so that the model can extract more feature information about wheat straw in the feature extraction stage; Secondly, after the improved C3 module is embedded the location attention module (Coordinate Attention, CA), by capturing the long-distance dependence on the space and the channel, makes the model more focused on the feature extraction of the target area, and further strengthens the feature extraction ability of the backbone network; Finally, for the traditional frame regression loss the function cannot solve the problem of returning gradients when the predicted frame and the real frame intersect. It is proposed to use CIoU instead of the traditional GIoU, and continue to guide the predicted frame while considering the Euclidean distance and aspect ratio of the center point of the predicted frame and the real frame. Moving closer to the ground-truth box, the loss function is further reduced. On the homemade wheat straw dataset, under the same training strategy, the experimental results show that! Compared with the traditional YOLOv5 model, the improved model has a 1.71% increase in mAP, which proves that the improved model is superior to the traditional YOLOv5 model in terms of accuracy, and has a better detection effect on small targets such as wheat straw some practicality.
{"title":"Wheat straw target detection algorithm based on improved YOLOv5","authors":"Pengfei Li, Heng Wang, Xueyu Huang","doi":"10.1109/ISPDS56360.2022.9874020","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874020","url":null,"abstract":"In agricultural production, the growth and yield of crops have always attracted people's attention. For the detection of wheat planting density, a wheat straw detection model based on improved YOLOv5 is proposed in this paper. Firstly, at the end of the backbone network, the C3 module (C3TR) integrated with Transformer is used to replace the traditional C3 module, so that the model can extract more feature information about wheat straw in the feature extraction stage; Secondly, after the improved C3 module is embedded the location attention module (Coordinate Attention, CA), by capturing the long-distance dependence on the space and the channel, makes the model more focused on the feature extraction of the target area, and further strengthens the feature extraction ability of the backbone network; Finally, for the traditional frame regression loss the function cannot solve the problem of returning gradients when the predicted frame and the real frame intersect. It is proposed to use CIoU instead of the traditional GIoU, and continue to guide the predicted frame while considering the Euclidean distance and aspect ratio of the center point of the predicted frame and the real frame. Moving closer to the ground-truth box, the loss function is further reduced. On the homemade wheat straw dataset, under the same training strategy, the experimental results show that! Compared with the traditional YOLOv5 model, the improved model has a 1.71% increase in mAP, which proves that the improved model is superior to the traditional YOLOv5 model in terms of accuracy, and has a better detection effect on small targets such as wheat straw some practicality.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121671976","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874130
X. Gao, Huijuan Song, Yan Li, Qing Zhao, Wei Li, Yingang Zhang, Lu Chao
The official abolition of 487 toll booths at provincial borders in 2020 marked the formal formation of the world's largest motorway toll collection network, creating a brand-new situation of “one network operation and integrated services” for motorways. It also put forward new requirements and challenges for the user services of the ETC. This paper addresses the problems existing in the voice quality inspection of ETC customer service network, such as “the scope of quality inspection is not wide” and “the efficiency of quality inspection is not high”. Based on AI technology, the logic of speech recognition, role recognition, semantic recognition and emotion recognition is established, and the intelligent quality inspection model is constructed. The operation data shows that these efforts effectively improve the quality of service efficiency and service quality and lay a foundation for steady and quality development of ETC user service.
{"title":"Research on Intelligent Quality Inspection of Customer Service Under the “One Network” Operation Mode of Toll Roads","authors":"X. Gao, Huijuan Song, Yan Li, Qing Zhao, Wei Li, Yingang Zhang, Lu Chao","doi":"10.1109/ISPDS56360.2022.9874130","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874130","url":null,"abstract":"The official abolition of 487 toll booths at provincial borders in 2020 marked the formal formation of the world's largest motorway toll collection network, creating a brand-new situation of “one network operation and integrated services” for motorways. It also put forward new requirements and challenges for the user services of the ETC. This paper addresses the problems existing in the voice quality inspection of ETC customer service network, such as “the scope of quality inspection is not wide” and “the efficiency of quality inspection is not high”. Based on AI technology, the logic of speech recognition, role recognition, semantic recognition and emotion recognition is established, and the intelligent quality inspection model is constructed. The operation data shows that these efforts effectively improve the quality of service efficiency and service quality and lay a foundation for steady and quality development of ETC user service.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"86 9","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"113940659","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874005
Ran Zhang, Yifan Wang, Yifeng Yin
As an important part of artificial intelligence technology, deep learning is widely used in various fields of contemporary society. The security of deep learning directly affects the effectiveness of its application in different fields. Effective attack algorithms can evaluate the security of deep learning models, and black box attacks are one of the important methods for testing the security of deep learning algorithms. An adaptive black box attack algorithm based on improved differential evolution is proposed to solve the problems of many queries, difficult selection of attack points that may cause higher attack costs in applications. The algorithm sets the variation factor as a linear decreasing function, uses the fitness function to adaptively control the change of the cross probability factor to improve the global search ability and accelerate the convergence rate, proposes a new variation strategy to enhance the ability of global search and local exploitation and the accuracy of searching attack points, and optimizes the loss function and the calculation method of gradient for defining decisions in deep learning models to improve the effectiveness and efficiency of black box attacks. The results of the comparison experiments show that the attack success rate is effectively improved and the average time and the average number of queries are reduced with the same attack success rate.
{"title":"An adaptive black box attack algorithm based on improved differential evolution","authors":"Ran Zhang, Yifan Wang, Yifeng Yin","doi":"10.1109/ISPDS56360.2022.9874005","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874005","url":null,"abstract":"As an important part of artificial intelligence technology, deep learning is widely used in various fields of contemporary society. The security of deep learning directly affects the effectiveness of its application in different fields. Effective attack algorithms can evaluate the security of deep learning models, and black box attacks are one of the important methods for testing the security of deep learning algorithms. An adaptive black box attack algorithm based on improved differential evolution is proposed to solve the problems of many queries, difficult selection of attack points that may cause higher attack costs in applications. The algorithm sets the variation factor as a linear decreasing function, uses the fitness function to adaptively control the change of the cross probability factor to improve the global search ability and accelerate the convergence rate, proposes a new variation strategy to enhance the ability of global search and local exploitation and the accuracy of searching attack points, and optimizes the loss function and the calculation method of gradient for defining decisions in deep learning models to improve the effectiveness and efficiency of black box attacks. The results of the comparison experiments show that the attack success rate is effectively improved and the average time and the average number of queries are reduced with the same attack success rate.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124169658","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：