Pub Date : 2023-03-01DOI: 10.1109/prmvia58252.2023.00042
Ying Zhang, Dongqiang Shi, Di Peng, Tiantian Li, Qian Zhao
The rocket has a variety of working modes after delivery, and according to its unique aircraft characteristics, this paper proposes an exponentially smooth rocket mode switching algorithm. The fixed delay caused by mode switching may also affect the startup time of normal working mode, so only enter shallow sleep mode when the idle time is short. When idle for a long time, the system needs to switch to deep sleep mode, where the power consumption of the system is minimized. The delay and power consumption of mode switching are analyzed and verified. After mathematical modeling, the smallest weight parameter of MRE is taken to meet the requirements of reliable and fast switching between multiple working modes of the rocket.
{"title":"Exponentially Smooth Rocket Mode Switching Algorithm Simulation","authors":"Ying Zhang, Dongqiang Shi, Di Peng, Tiantian Li, Qian Zhao","doi":"10.1109/prmvia58252.2023.00042","DOIUrl":"https://doi.org/10.1109/prmvia58252.2023.00042","url":null,"abstract":"The rocket has a variety of working modes after delivery, and according to its unique aircraft characteristics, this paper proposes an exponentially smooth rocket mode switching algorithm. The fixed delay caused by mode switching may also affect the startup time of normal working mode, so only enter shallow sleep mode when the idle time is short. When idle for a long time, the system needs to switch to deep sleep mode, where the power consumption of the system is minimized. The delay and power consumption of mode switching are analyzed and verified. After mathematical modeling, the smallest weight parameter of MRE is taken to meet the requirements of reliable and fast switching between multiple working modes of the rocket.","PeriodicalId":221346,"journal":{"name":"2023 International Conference on Pattern Recognition, Machine Vision and Intelligent Algorithms (PRMVIA)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124910782","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 : 2023-03-01DOI: 10.1109/prmvia58252.2023.00026
Xiaobo Yu
Artificial intelligence is the crystallization of human intelligence and an important transformation of human labor tools. In the future development process, artificial intelligence will become a new and high-quality tool, which will bring great value to mankind. The engraving machine numerical control (CNC) system is an automatic tracking system, which has strong practical value. According to the characteristics of NURBS interpolation algorithm, this paper designs a CNC sculpture control system based on NURBS interpolation algorithm, and improves its dynamic characteristics and positioning accuracy. Through the system performance test of the CNC sculptor designed in this paper, starting from the stability and safety, we can see from the experimental data that the CNC sculptor designed in this paper has high stability and safety. In addition, compared with the time spent by the market engraving machine and the CNC engraving machine designed in this paper, it takes 22 minutes and 20 minutes for the market engraving machine and the CNC engraving machine designed in this paper to carve one wooden plate respectively. To sum up, the CNC sculpture control system designed in this paper has a good effect and is worthy of further promotion and application.
{"title":"Design and Implementation of CNC Sculpture Control System Based on NURBS Interpolation Algorithm","authors":"Xiaobo Yu","doi":"10.1109/prmvia58252.2023.00026","DOIUrl":"https://doi.org/10.1109/prmvia58252.2023.00026","url":null,"abstract":"Artificial intelligence is the crystallization of human intelligence and an important transformation of human labor tools. In the future development process, artificial intelligence will become a new and high-quality tool, which will bring great value to mankind. The engraving machine numerical control (CNC) system is an automatic tracking system, which has strong practical value. According to the characteristics of NURBS interpolation algorithm, this paper designs a CNC sculpture control system based on NURBS interpolation algorithm, and improves its dynamic characteristics and positioning accuracy. Through the system performance test of the CNC sculptor designed in this paper, starting from the stability and safety, we can see from the experimental data that the CNC sculptor designed in this paper has high stability and safety. In addition, compared with the time spent by the market engraving machine and the CNC engraving machine designed in this paper, it takes 22 minutes and 20 minutes for the market engraving machine and the CNC engraving machine designed in this paper to carve one wooden plate respectively. To sum up, the CNC sculpture control system designed in this paper has a good effect and is worthy of further promotion and application.","PeriodicalId":221346,"journal":{"name":"2023 International Conference on Pattern Recognition, Machine Vision and Intelligent Algorithms (PRMVIA)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123485316","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 : 2023-03-01DOI: 10.1109/prmvia58252.2023.00022
T. Yu, Lei Li, Xunlian Luo, Qiang Li
In the scene of equipment maintenance, the equipment parts target detection technology can provide technical support for maintenance personnel, and lightweight algorithms based on deep learning have been much concerned, which have the advantages of strong feature extraction and short delay time. YOLOv7 is considered as a new algorithm in the YOLO series, which offers many optimized modules to improve target detection abilities. However, YOLOv7 has problems such as huge amount of computation and parameters, serious memory consumption, and the over-optimized structure. In this paper, a lightweight algorithm ReM-YOLO based on YOLOv7 is proposed to improve the network structure. YOLOv7 is improved by adding C3 blocks, MobileOne blocks and Rep-DSC blocks to reduce the model size while maintaining high precision, and a non-parameter SimAM attention module is employed to further improve the detection accuracy. Compared to YOLOv7, the ReM-YOLO has better improvements in precision and recall, and the model size is reduced by 1/3 size of YOLOv7. It has been observed that experimental tests are carried out on our dataset of vehicle engine components with the high accuracy rate of 96.2%. The improved algorithm helps further experiments about model compression effectively.
{"title":"ReM-YOLO: A New Lightweight Vehicle Parts Target Detection Algorithm","authors":"T. Yu, Lei Li, Xunlian Luo, Qiang Li","doi":"10.1109/prmvia58252.2023.00022","DOIUrl":"https://doi.org/10.1109/prmvia58252.2023.00022","url":null,"abstract":"In the scene of equipment maintenance, the equipment parts target detection technology can provide technical support for maintenance personnel, and lightweight algorithms based on deep learning have been much concerned, which have the advantages of strong feature extraction and short delay time. YOLOv7 is considered as a new algorithm in the YOLO series, which offers many optimized modules to improve target detection abilities. However, YOLOv7 has problems such as huge amount of computation and parameters, serious memory consumption, and the over-optimized structure. In this paper, a lightweight algorithm ReM-YOLO based on YOLOv7 is proposed to improve the network structure. YOLOv7 is improved by adding C3 blocks, MobileOne blocks and Rep-DSC blocks to reduce the model size while maintaining high precision, and a non-parameter SimAM attention module is employed to further improve the detection accuracy. Compared to YOLOv7, the ReM-YOLO has better improvements in precision and recall, and the model size is reduced by 1/3 size of YOLOv7. It has been observed that experimental tests are carried out on our dataset of vehicle engine components with the high accuracy rate of 96.2%. The improved algorithm helps further experiments about model compression effectively.","PeriodicalId":221346,"journal":{"name":"2023 International Conference on Pattern Recognition, Machine Vision and Intelligent Algorithms (PRMVIA)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116314302","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}
Zero-Shot Learning (ZSL) simulates human’s transfer learning mechanism, which can recognize samples or categories that have not appeared during the training phase. However, the current ZSL still has a domain shift issue. To solved the domain shift issue, we propose a new ZSL method that combines Vision Transformer (ViT) and the encoder-decoder mechanism. This method refers to ViT’s Multi-Head Self-Attention (MSA) to extract more detailed visual features. The encoder-decoder mechanism can make the semantic information extracted from the image features accurately express its visual features and enhance recognition accuracy. We implemented it on three data sets of CUB, SUN and AWA2, and the experimental results proved that the method suggested in this study performs better than the current available methods. It shows that our new method is an effective ZSL method.
{"title":"Zero-Shot Learning based on Vision Transformer","authors":"Ruisheng Ran, Qianwei Hu, Tianyu Gao, Shuhong Dong","doi":"10.1109/prmvia58252.2023.00010","DOIUrl":"https://doi.org/10.1109/prmvia58252.2023.00010","url":null,"abstract":"Zero-Shot Learning (ZSL) simulates human’s transfer learning mechanism, which can recognize samples or categories that have not appeared during the training phase. However, the current ZSL still has a domain shift issue. To solved the domain shift issue, we propose a new ZSL method that combines Vision Transformer (ViT) and the encoder-decoder mechanism. This method refers to ViT’s Multi-Head Self-Attention (MSA) to extract more detailed visual features. The encoder-decoder mechanism can make the semantic information extracted from the image features accurately express its visual features and enhance recognition accuracy. We implemented it on three data sets of CUB, SUN and AWA2, and the experimental results proved that the method suggested in this study performs better than the current available methods. It shows that our new method is an effective ZSL method.","PeriodicalId":221346,"journal":{"name":"2023 International Conference on Pattern Recognition, Machine Vision and Intelligent Algorithms (PRMVIA)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129320714","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 : 2023-03-01DOI: 10.1109/PRMVIA58252.2023.00020
Jingjie Yang, Xiang Zheng
Partial discharge (PD) monitor in gas-insulated switchgear (GIS) is an important means to detect insulation defects of equipment. To solve the problem that the traditional PD extraction features are not obvious and recognition precision is limited. The paper presents a new pattern recognition algorithm by combining the multiscale dispersion entropy (MDE), locally linear embedding (LLE), and stacking ensemble learning, to effectively refine the recognition correct rate of PD types. First, the MDE values of PD signal were calculated as the feature value. Then, use LLE to reduce dimensions to refine the speed and precision of model recognition. Finally, use stacking ensemble learning to train and recognize the feature values after dimension reduction. Among them, K-nearest neighbor, random forest and Gaussian Bayes were selected for the first layer learners, and logical regression model was selected for the second layer learner. The validation results indicated that the recognition correct rate of the proposed algorithm for four typical PD types in GIS was more than 98%, and it has a strong anti-interference ability, which is significantly better than the traditional feature extraction methods.
{"title":"Partial Discharge Pattern Recognition in GIS Based on Multiscale Dispersion Entropy and Stacking Ensemble Learning","authors":"Jingjie Yang, Xiang Zheng","doi":"10.1109/PRMVIA58252.2023.00020","DOIUrl":"https://doi.org/10.1109/PRMVIA58252.2023.00020","url":null,"abstract":"Partial discharge (PD) monitor in gas-insulated switchgear (GIS) is an important means to detect insulation defects of equipment. To solve the problem that the traditional PD extraction features are not obvious and recognition precision is limited. The paper presents a new pattern recognition algorithm by combining the multiscale dispersion entropy (MDE), locally linear embedding (LLE), and stacking ensemble learning, to effectively refine the recognition correct rate of PD types. First, the MDE values of PD signal were calculated as the feature value. Then, use LLE to reduce dimensions to refine the speed and precision of model recognition. Finally, use stacking ensemble learning to train and recognize the feature values after dimension reduction. Among them, K-nearest neighbor, random forest and Gaussian Bayes were selected for the first layer learners, and logical regression model was selected for the second layer learner. The validation results indicated that the recognition correct rate of the proposed algorithm for four typical PD types in GIS was more than 98%, and it has a strong anti-interference ability, which is significantly better than the traditional feature extraction methods.","PeriodicalId":221346,"journal":{"name":"2023 International Conference on Pattern Recognition, Machine Vision and Intelligent Algorithms (PRMVIA)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125702221","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 : 2023-03-01DOI: 10.1109/prmvia58252.2023.00047
Jing Chen, Guan Yang, Xiaoming Liu, Yang Liu, Weifu Chen
Aiming at the existing problems in the few-shot learning methods which treat samples in an isolated perspective and ignore the difference information between samples, we propose a few-shot learning classification network based on feature differentiation. A new feature adaptive fusion module and a feature conversion module form our network, where the former is proposed to fuse global information and detailed features, and the latter one marks the semantic features which have high recognition, so as to narrow the semantics within the same category and widen the semantic gap between different categories. CUB dataset and mini-ImageNet dataset were used in the experiment, and the accuracy of 5way-lshot and 5way-5shot tasks respectively achieved 57.63%, 76.54% and 54.39%, 73.19%. Experimental results show that our method can further learn how to distinguish different category concepts through differentiated features, thus the proposed few-shot learning model has higher accuracy and robustness.
{"title":"Few-shot Classification Network Based on Feature Differentiation","authors":"Jing Chen, Guan Yang, Xiaoming Liu, Yang Liu, Weifu Chen","doi":"10.1109/prmvia58252.2023.00047","DOIUrl":"https://doi.org/10.1109/prmvia58252.2023.00047","url":null,"abstract":"Aiming at the existing problems in the few-shot learning methods which treat samples in an isolated perspective and ignore the difference information between samples, we propose a few-shot learning classification network based on feature differentiation. A new feature adaptive fusion module and a feature conversion module form our network, where the former is proposed to fuse global information and detailed features, and the latter one marks the semantic features which have high recognition, so as to narrow the semantics within the same category and widen the semantic gap between different categories. CUB dataset and mini-ImageNet dataset were used in the experiment, and the accuracy of 5way-lshot and 5way-5shot tasks respectively achieved 57.63%, 76.54% and 54.39%, 73.19%. Experimental results show that our method can further learn how to distinguish different category concepts through differentiated features, thus the proposed few-shot learning model has higher accuracy and robustness.","PeriodicalId":221346,"journal":{"name":"2023 International Conference on Pattern Recognition, Machine Vision and Intelligent Algorithms (PRMVIA)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130115935","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 : 2023-03-01DOI: 10.1109/PRMVIA58252.2023.00039
C. Peng, Bing He, Wenqiang Xi, Guancheng Lin
Polar harmonic Fourier moments (PHFMs) are popular for image analysis due to their properties of lower computation complexity and minimal redundant description capability of images. However, the traditional PHFMs are unavailable for color stereo image analysis on the one hand, and on the other hand the polar harmonic polynomials with integer-order are not able to extract fine features. In this paper, a new category of moments named octonion fractional-order PHFMs (OFrPHFMs) are proposed using the fractional-order basis functions of PHFMs and octonion theory. The proposed moments can be viewed as a generalization of quaternion orthogonal moments. Furthermore, since the image moments formed by the octonion descriptor can treat the color stereo image integrally, it has a strong representation capability. More importantly, some numerical instability and calculation issues are discussed and a fast computational framework using matrix operation and block Gaussian numerical integration is developed to improve the accuracy and efficiency of the proposed OFrPHFMs. Finally, to demonstrate the validation of the proposed moments, the image experiments are conducted and the results show that the proposed OFrPHFMs have favorable performance in the field of color stereo image analysis.
{"title":"Stereo Image Analysis by Octonion Fractional-Order Orthogonal Color Moments","authors":"C. Peng, Bing He, Wenqiang Xi, Guancheng Lin","doi":"10.1109/PRMVIA58252.2023.00039","DOIUrl":"https://doi.org/10.1109/PRMVIA58252.2023.00039","url":null,"abstract":"Polar harmonic Fourier moments (PHFMs) are popular for image analysis due to their properties of lower computation complexity and minimal redundant description capability of images. However, the traditional PHFMs are unavailable for color stereo image analysis on the one hand, and on the other hand the polar harmonic polynomials with integer-order are not able to extract fine features. In this paper, a new category of moments named octonion fractional-order PHFMs (OFrPHFMs) are proposed using the fractional-order basis functions of PHFMs and octonion theory. The proposed moments can be viewed as a generalization of quaternion orthogonal moments. Furthermore, since the image moments formed by the octonion descriptor can treat the color stereo image integrally, it has a strong representation capability. More importantly, some numerical instability and calculation issues are discussed and a fast computational framework using matrix operation and block Gaussian numerical integration is developed to improve the accuracy and efficiency of the proposed OFrPHFMs. Finally, to demonstrate the validation of the proposed moments, the image experiments are conducted and the results show that the proposed OFrPHFMs have favorable performance in the field of color stereo image analysis.","PeriodicalId":221346,"journal":{"name":"2023 International Conference on Pattern Recognition, Machine Vision and Intelligent Algorithms (PRMVIA)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131575001","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 : 2023-03-01DOI: 10.1109/prmvia58252.2023.00044
Guojun Wang, Zhenzhen Li, Jianbin Yao
In the aspect of low-contrast feature extraction and registration of concrete structure surface, traditional algorithms have some problems such as low computational efficiency, less feature extraction and low matching accuracy. The method based on deep learning has become a mainstream method at present, but the supervised learning method based on manual annotation has the problem that low contrast features cannot be marked. In view of this, it is necessary to study the most promising deep learning method based on graph convolution for progressive extraction and registration of low-contrast features of concrete structure surfaces. This paper uses Superpoint framework to solve the low contrast problem at the end of supervised learning. The indoor and outdoor test results show that the deflection curve trend of measuring points is basically consistent with that of the displacement meter, which indicates the robustness of feature point tracking based on SuperGlue, and further indicates that the method can be used as an effective technical reserve for deflection measurement of concrete structures.
{"title":"Research on Low Contrast Feature Extraction and Registration Effect of Concrete Structure based on SuperGlue Algorithm","authors":"Guojun Wang, Zhenzhen Li, Jianbin Yao","doi":"10.1109/prmvia58252.2023.00044","DOIUrl":"https://doi.org/10.1109/prmvia58252.2023.00044","url":null,"abstract":"In the aspect of low-contrast feature extraction and registration of concrete structure surface, traditional algorithms have some problems such as low computational efficiency, less feature extraction and low matching accuracy. The method based on deep learning has become a mainstream method at present, but the supervised learning method based on manual annotation has the problem that low contrast features cannot be marked. In view of this, it is necessary to study the most promising deep learning method based on graph convolution for progressive extraction and registration of low-contrast features of concrete structure surfaces. This paper uses Superpoint framework to solve the low contrast problem at the end of supervised learning. The indoor and outdoor test results show that the deflection curve trend of measuring points is basically consistent with that of the displacement meter, which indicates the robustness of feature point tracking based on SuperGlue, and further indicates that the method can be used as an effective technical reserve for deflection measurement of concrete structures.","PeriodicalId":221346,"journal":{"name":"2023 International Conference on Pattern Recognition, Machine Vision and Intelligent Algorithms (PRMVIA)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127580392","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 : 2023-03-01DOI: 10.1109/PRMVIA58252.2023.00036
Chuancheng Deng, Feng Yi, Xing Li, Junyi Tang, Guang Sun
Decision tree algorithm belongs to the machine learning method of supervised learning, and it is one of the common techniques of data mining. CHAID algorithm is one of the common decision tree algorithms. The CHAID algorithm can be used for prediction as well as for classification. It is mainly used in market analysis, risk prediction and other fields. The second part tells the core idea of CHAID algorithm and the specific steps of the classification process, as well as the principle formula of chi-square detection and the specific steps of the calculation mainly used by CHAID algorithm. The third part describes the branch principle of CHAID algorithm and other decision tree algorithm, and analyses the accuracy of CHAID algorithm. Let us choose a relatively good decision tree algorithm according to the specific data situation, and apply the CHAID algorithm can also make some countermeasures according to some factors affecting the accuracy, which is more conducive to get a more accurate and better result.
{"title":"Performance Analysis of CHAID Algorithm for Accuracy","authors":"Chuancheng Deng, Feng Yi, Xing Li, Junyi Tang, Guang Sun","doi":"10.1109/PRMVIA58252.2023.00036","DOIUrl":"https://doi.org/10.1109/PRMVIA58252.2023.00036","url":null,"abstract":"Decision tree algorithm belongs to the machine learning method of supervised learning, and it is one of the common techniques of data mining. CHAID algorithm is one of the common decision tree algorithms. The CHAID algorithm can be used for prediction as well as for classification. It is mainly used in market analysis, risk prediction and other fields. The second part tells the core idea of CHAID algorithm and the specific steps of the classification process, as well as the principle formula of chi-square detection and the specific steps of the calculation mainly used by CHAID algorithm. The third part describes the branch principle of CHAID algorithm and other decision tree algorithm, and analyses the accuracy of CHAID algorithm. Let us choose a relatively good decision tree algorithm according to the specific data situation, and apply the CHAID algorithm can also make some countermeasures according to some factors affecting the accuracy, which is more conducive to get a more accurate and better result.","PeriodicalId":221346,"journal":{"name":"2023 International Conference on Pattern Recognition, Machine Vision and Intelligent Algorithms (PRMVIA)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115621306","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 : 2023-03-01DOI: 10.1109/prmvia58252.2023.00012
Feiyang Qin, Wenqi Na, Song Gao, Shaowen Yao
Although deep neural networks (DNNs) have achieved exceptional performance, they are shown to be fragile to universal adversarial perturbations (UAP), which can be applied to any images to fool a well-trained DNN. Several methods have been proposed to design universal perturbations. However, these methods often leave visible traces in natural images. In this paper, we propose Sigma-UAP, a semi-universal adversarial attack, to enhance the quasi-imperceptibility of universal adversarial perturbations, in which the Sigma-map algorithm is leveraged to hide perturbations by identifying the low-frequency region of the image and eliminating the perturbations in that region. Then, we use a simple matrix calculation to augment the perturbation in the high-frequency region to ensure the attack effectiveness of the perturbation. The extensive empirical experiments show that, compared with the state-of-the-art universal adversarial attacks, Sigma-UAP method obtains excellent attack capabilities in visual effect and attack success rate.
{"title":"Sigma-UAP: An Invisible Semi-Universal Adversarial Attack Against Deep Neural Networks","authors":"Feiyang Qin, Wenqi Na, Song Gao, Shaowen Yao","doi":"10.1109/prmvia58252.2023.00012","DOIUrl":"https://doi.org/10.1109/prmvia58252.2023.00012","url":null,"abstract":"Although deep neural networks (DNNs) have achieved exceptional performance, they are shown to be fragile to universal adversarial perturbations (UAP), which can be applied to any images to fool a well-trained DNN. Several methods have been proposed to design universal perturbations. However, these methods often leave visible traces in natural images. In this paper, we propose Sigma-UAP, a semi-universal adversarial attack, to enhance the quasi-imperceptibility of universal adversarial perturbations, in which the Sigma-map algorithm is leveraged to hide perturbations by identifying the low-frequency region of the image and eliminating the perturbations in that region. Then, we use a simple matrix calculation to augment the perturbation in the high-frequency region to ensure the attack effectiveness of the perturbation. The extensive empirical experiments show that, compared with the state-of-the-art universal adversarial attacks, Sigma-UAP method obtains excellent attack capabilities in visual effect and attack success rate.","PeriodicalId":221346,"journal":{"name":"2023 International Conference on Pattern Recognition, Machine Vision and Intelligent Algorithms (PRMVIA)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114068438","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：