Pub Date : 2021-11-22DOI: 10.1109/ICESIT53460.2021.9696758
Zhan Yuzhuo
Since the 21st century, the vigorous development of big data technology has led to its application in the electrical power systems, while some initial progress has been made in the research on the application of big data technology in the electric power system. This paper analyzes the application of big data technology in electric power system from the development status of big data technology as well as electric power system. The key technologies of the application of big data technology in the electric power system are divided into integration and management technology, data processing technology, data analysis technology and data visualization technology of electric power big data, and analyzed one by one. Meanwhile, this paper also lists the application examples of electric power mega data technology in smart grid, so as to confirm the development trend of electric power system under big data technology.
{"title":"A Study on Power System Development Trend through Comptuer Visualization and Big Data Technology","authors":"Zhan Yuzhuo","doi":"10.1109/ICESIT53460.2021.9696758","DOIUrl":"https://doi.org/10.1109/ICESIT53460.2021.9696758","url":null,"abstract":"Since the 21st century, the vigorous development of big data technology has led to its application in the electrical power systems, while some initial progress has been made in the research on the application of big data technology in the electric power system. This paper analyzes the application of big data technology in electric power system from the development status of big data technology as well as electric power system. The key technologies of the application of big data technology in the electric power system are divided into integration and management technology, data processing technology, data analysis technology and data visualization technology of electric power big data, and analyzed one by one. Meanwhile, this paper also lists the application examples of electric power mega data technology in smart grid, so as to confirm the development trend of electric power system under big data technology.","PeriodicalId":164745,"journal":{"name":"2021 IEEE International Conference on Emergency Science and Information Technology (ICESIT)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121050172","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 : 2021-11-22DOI: 10.1109/ICESIT53460.2021.9696759
Biao Li, Jing Liu
With the increasing contradiction between less people and more stations and the development of front-end state sensing technology, data transmission technology and intelligent diagnosis technology, in order to effectively lighten the work strength of operation and maintenance staff and improve the operation and maintenance efficiency, we put forward the construction scheme of remote intelligent management platform, which makes use of modern information technology and advanced communication technology like the existing mobile Internet and artificial intelligence to realizes “Internet of Everything” and human-computer interaction at all stations under the jurisdiction, so as to make it an intelligent service system featuring comprehensive state perception, efficient information processing and convenient and flexible application.
{"title":"Research on Remote Intelligent Platform and Automatic Monitoring System of Transformer Substations","authors":"Biao Li, Jing Liu","doi":"10.1109/ICESIT53460.2021.9696759","DOIUrl":"https://doi.org/10.1109/ICESIT53460.2021.9696759","url":null,"abstract":"With the increasing contradiction between less people and more stations and the development of front-end state sensing technology, data transmission technology and intelligent diagnosis technology, in order to effectively lighten the work strength of operation and maintenance staff and improve the operation and maintenance efficiency, we put forward the construction scheme of remote intelligent management platform, which makes use of modern information technology and advanced communication technology like the existing mobile Internet and artificial intelligence to realizes “Internet of Everything” and human-computer interaction at all stations under the jurisdiction, so as to make it an intelligent service system featuring comprehensive state perception, efficient information processing and convenient and flexible application.","PeriodicalId":164745,"journal":{"name":"2021 IEEE International Conference on Emergency Science and Information Technology (ICESIT)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127829074","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 social model is a huge virtual platform where to freely express themselves and give their views and feelings, influencing any aspect of life, with implications for marketing and communication alike. Aspect words sentiment analysis can more accurately understand user needs and improve enterprise marketing strategies. In current researches on aspect words sentiment analysis, researchers use the integration of attention mechanism and LSTM to obtain key information. However, there are few studies on the fusion of aspect words, context, and multi-layer deep memory networks. Therefore, we proposed a multi-layer deep memory network model based on the splicing of aspect terms and context vectors. The model can further strengthen the fusion between aspect words and context vectors, and make up for the shortcomings of LSTM in transmitting information loss. Experimental results on Restaurant and Laptop datasets show that the proposed method has a better performance.
{"title":"Aspect-words Sentiment analysis of commodity comments based on deep memory network","authors":"Wenjun Cheng, Jike Ge, Chengzhi Wu, Sheng Yu, Haoyin Liu, Jichao Xu","doi":"10.1109/ICESIT53460.2021.9696708","DOIUrl":"https://doi.org/10.1109/ICESIT53460.2021.9696708","url":null,"abstract":"The social model is a huge virtual platform where to freely express themselves and give their views and feelings, influencing any aspect of life, with implications for marketing and communication alike. Aspect words sentiment analysis can more accurately understand user needs and improve enterprise marketing strategies. In current researches on aspect words sentiment analysis, researchers use the integration of attention mechanism and LSTM to obtain key information. However, there are few studies on the fusion of aspect words, context, and multi-layer deep memory networks. Therefore, we proposed a multi-layer deep memory network model based on the splicing of aspect terms and context vectors. The model can further strengthen the fusion between aspect words and context vectors, and make up for the shortcomings of LSTM in transmitting information loss. Experimental results on Restaurant and Laptop datasets show that the proposed method has a better performance.","PeriodicalId":164745,"journal":{"name":"2021 IEEE International Conference on Emergency Science and Information Technology (ICESIT)","volume":"158 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133055302","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 : 2021-11-22DOI: 10.1109/ICESIT53460.2021.9696936
Jing Li, Bin Zhang, Haiqing Li
The cutter is one critical component in a milling tool, and its operational condition directly affects the part machining quality and production efficiency. In this paper, a new method for milling cutters health monitoring is proposed. The proposed method extracts nonlinear entropy features with adaptive decomposition of the original multi-sensor monitoring signals. Then the extracted features are selected and adaptively fused into a virtual health indicator (HI) by self-organizing mapping (SOM) network to characterize the operational health condition of the milling cutter. High speed milling data from 2010 prognostics and health management (PHM) challenge is studied to demonstrate performance of the presented method. Experimental results show that the approach can effectively integrate the online multi-sensor signals to reliably describe health degradation of the milling cutter.
{"title":"Health Monitoring of Milling Cutters with Nonlinear Entropy and Self-organizing Mapping","authors":"Jing Li, Bin Zhang, Haiqing Li","doi":"10.1109/ICESIT53460.2021.9696936","DOIUrl":"https://doi.org/10.1109/ICESIT53460.2021.9696936","url":null,"abstract":"The cutter is one critical component in a milling tool, and its operational condition directly affects the part machining quality and production efficiency. In this paper, a new method for milling cutters health monitoring is proposed. The proposed method extracts nonlinear entropy features with adaptive decomposition of the original multi-sensor monitoring signals. Then the extracted features are selected and adaptively fused into a virtual health indicator (HI) by self-organizing mapping (SOM) network to characterize the operational health condition of the milling cutter. High speed milling data from 2010 prognostics and health management (PHM) challenge is studied to demonstrate performance of the presented method. Experimental results show that the approach can effectively integrate the online multi-sensor signals to reliably describe health degradation of the milling cutter.","PeriodicalId":164745,"journal":{"name":"2021 IEEE International Conference on Emergency Science and Information Technology (ICESIT)","volume":"80 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133784666","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 : 2021-11-22DOI: 10.1109/ICESIT53460.2021.9696720
Liu Jiahong, Bao Jie, Chen Yingshuang, Lv Chun
In this paper, a speaker recognition strategy in military radio communication is applied. In military operations, the most commonly used method of information transmission is radio communication. Speaker recognition technology can confirm the sender's identity, and effectively prevent the enemy from pretending to be our military commander to issue false orders. However, the datasets of the military commander from the radio are confidential, and there are no large open-source datasets. Consequently, speaker recognition accuracy is not ideal if we only train a small sample of speaker datasets. Therefore, we propose a transfer learning method for training. We pre-train a Deep Residual neural network (ResNet) with large sample datasets and re-train a novel adaptive model with a simple sample dataset in radio communication. Experiments are carried out using the aishell-2 dataset and the self-collected radio military command datasets. Experimental results demonstrate that the adaptive network with transfer learning method improves the performance by 23.55% relatively compared to the baseline system in radio communication.
{"title":"An Adaptive ResNet Based Speaker Recognition in Radio Communication","authors":"Liu Jiahong, Bao Jie, Chen Yingshuang, Lv Chun","doi":"10.1109/ICESIT53460.2021.9696720","DOIUrl":"https://doi.org/10.1109/ICESIT53460.2021.9696720","url":null,"abstract":"In this paper, a speaker recognition strategy in military radio communication is applied. In military operations, the most commonly used method of information transmission is radio communication. Speaker recognition technology can confirm the sender's identity, and effectively prevent the enemy from pretending to be our military commander to issue false orders. However, the datasets of the military commander from the radio are confidential, and there are no large open-source datasets. Consequently, speaker recognition accuracy is not ideal if we only train a small sample of speaker datasets. Therefore, we propose a transfer learning method for training. We pre-train a Deep Residual neural network (ResNet) with large sample datasets and re-train a novel adaptive model with a simple sample dataset in radio communication. Experiments are carried out using the aishell-2 dataset and the self-collected radio military command datasets. Experimental results demonstrate that the adaptive network with transfer learning method improves the performance by 23.55% relatively compared to the baseline system in radio communication.","PeriodicalId":164745,"journal":{"name":"2021 IEEE International Conference on Emergency Science and Information Technology (ICESIT)","volume":"59 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134078259","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 : 2021-11-22DOI: 10.1109/ICESIT53460.2021.9696640
Fang Binhao, Huang Hong, Zhou Ziyun
Detecting abnormal traffic in real life often requires analyzing massive data (high-dimensional data) and unbalanced data. Aiming at the above problems, an intrusion detection model (SMBR-XGBDT) based on the combination of SMOTE algorithm and Boruta algorithm with Extreme Gradient Boosting (XGBoost) algorithm is proposed. The experiment selected 14367 extremely unbalanced samples based on the CIRA-CIC-DoHBrw-2020 data set, and detected 4 categories: DOH, Non-DoH, Benign-DoH, Malicious-DoH, using decision tree algorithm, random forest Algorithm, XGBoost algorithm as a control. The experimental results show that the SMBR-XGBDT model is significantly better than the other three models. The precision, recall, and F1 scores of the overall test were 93%, 93 %, and 93 %, respectively, which verified the effectiveness of the method. The precision rates of DOH, Non-DoH, Benign-DoH, Malicious-DoH were 88%, 100%, 98%, and 87%, respectively, which verified the feasibility of the method to deal with unbalanced data.
{"title":"Improve the Application of XGBDT in Network Abnormal Traffic Detection","authors":"Fang Binhao, Huang Hong, Zhou Ziyun","doi":"10.1109/ICESIT53460.2021.9696640","DOIUrl":"https://doi.org/10.1109/ICESIT53460.2021.9696640","url":null,"abstract":"Detecting abnormal traffic in real life often requires analyzing massive data (high-dimensional data) and unbalanced data. Aiming at the above problems, an intrusion detection model (SMBR-XGBDT) based on the combination of SMOTE algorithm and Boruta algorithm with Extreme Gradient Boosting (XGBoost) algorithm is proposed. The experiment selected 14367 extremely unbalanced samples based on the CIRA-CIC-DoHBrw-2020 data set, and detected 4 categories: DOH, Non-DoH, Benign-DoH, Malicious-DoH, using decision tree algorithm, random forest Algorithm, XGBoost algorithm as a control. The experimental results show that the SMBR-XGBDT model is significantly better than the other three models. The precision, recall, and F1 scores of the overall test were 93%, 93 %, and 93 %, respectively, which verified the effectiveness of the method. The precision rates of DOH, Non-DoH, Benign-DoH, Malicious-DoH were 88%, 100%, 98%, and 87%, respectively, which verified the feasibility of the method to deal with unbalanced data.","PeriodicalId":164745,"journal":{"name":"2021 IEEE International Conference on Emergency Science and Information Technology (ICESIT)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115038013","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 : 2021-11-22DOI: 10.1109/ICESIT53460.2021.9696835
Liwei Liu, Daming Qu, Alin Hou
The accurate segmentation of the lesion area is of great significance to the actual medical treatment. However, the segmentation results of the current segmentation network are not accurate enough to provide guidance for actual medical treatment. To solve this problem, a improved U-Net segmentation network is proposed. Firstly. The residual module and new attention mechanism are introduced to optimize the encoder, and 2×2 convolution is used instead of pooling operation, which can refine and extract features while retaining spatial feature information. Secondly, the attention mechanism is introduced before the upsampling jump connection, so that the network pays attention to the spatial information of the low-level feature map. The improved U-Net segmentation network was evaluated on the LiTS datasets. Compared with the traditional If-Net, the Dice coefficient and recall rate are increased by 5.6% and 3.03 % respectively in the liver segmentation task, the Dice coefficient and recall rate are increased by 7.51% and 8.8% respectively in the liver tumor segmentation task.
{"title":"A semantic segmentation algorithm supported by image processing and neural network","authors":"Liwei Liu, Daming Qu, Alin Hou","doi":"10.1109/ICESIT53460.2021.9696835","DOIUrl":"https://doi.org/10.1109/ICESIT53460.2021.9696835","url":null,"abstract":"The accurate segmentation of the lesion area is of great significance to the actual medical treatment. However, the segmentation results of the current segmentation network are not accurate enough to provide guidance for actual medical treatment. To solve this problem, a improved U-Net segmentation network is proposed. Firstly. The residual module and new attention mechanism are introduced to optimize the encoder, and 2×2 convolution is used instead of pooling operation, which can refine and extract features while retaining spatial feature information. Secondly, the attention mechanism is introduced before the upsampling jump connection, so that the network pays attention to the spatial information of the low-level feature map. The improved U-Net segmentation network was evaluated on the LiTS datasets. Compared with the traditional If-Net, the Dice coefficient and recall rate are increased by 5.6% and 3.03 % respectively in the liver segmentation task, the Dice coefficient and recall rate are increased by 7.51% and 8.8% respectively in the liver tumor segmentation task.","PeriodicalId":164745,"journal":{"name":"2021 IEEE International Conference on Emergency Science and Information Technology (ICESIT)","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115656707","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 : 2021-11-22DOI: 10.1109/ICESIT53460.2021.9696674
C. Hao, Yi Liming, Xu Zhengtao, Fang Qing
In the design of 500kV ultra-high voltage transmission towers in mountainous areas, the choice of tower slope will affect the size of the tower's root opening, thereby affecting the applicability of the tower in mountainous areas. At the same time, the choice of tower slope will also cause the tower weight and foundation force to be generated. Differences, thereby affecting the construction cost of the project. Combining actual engineering, taking a 20mm ice zone 53ZBC33 tower type for a 500kV line project as an example, a total of 8 different slopes of 0.09~0.16 for transmission towers of 60m, 54m, 48m, and 42m are calculated. The tower is used as an example. Steel quantity cost, foundation cost, and tower foundation land compensation fee are used as indicators to compare and analyze the project cost. When the calling height of the transmission tower is different, the corresponding economic slope is slightly different, but the overall economic slope fluctuates in the range of 0.10~0.13. In the actual design process, the economic slope of the tower can be quickly calculated in the slope range of 0.10~0.13.
{"title":"Study on computer modeling and calculation of 500kV ultra-high voltage transmission towers","authors":"C. Hao, Yi Liming, Xu Zhengtao, Fang Qing","doi":"10.1109/ICESIT53460.2021.9696674","DOIUrl":"https://doi.org/10.1109/ICESIT53460.2021.9696674","url":null,"abstract":"In the design of 500kV ultra-high voltage transmission towers in mountainous areas, the choice of tower slope will affect the size of the tower's root opening, thereby affecting the applicability of the tower in mountainous areas. At the same time, the choice of tower slope will also cause the tower weight and foundation force to be generated. Differences, thereby affecting the construction cost of the project. Combining actual engineering, taking a 20mm ice zone 53ZBC33 tower type for a 500kV line project as an example, a total of 8 different slopes of 0.09~0.16 for transmission towers of 60m, 54m, 48m, and 42m are calculated. The tower is used as an example. Steel quantity cost, foundation cost, and tower foundation land compensation fee are used as indicators to compare and analyze the project cost. When the calling height of the transmission tower is different, the corresponding economic slope is slightly different, but the overall economic slope fluctuates in the range of 0.10~0.13. In the actual design process, the economic slope of the tower can be quickly calculated in the slope range of 0.10~0.13.","PeriodicalId":164745,"journal":{"name":"2021 IEEE International Conference on Emergency Science and Information Technology (ICESIT)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115670626","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 : 2021-11-22DOI: 10.1109/ICESIT53460.2021.9696652
Xin Xia, Chuanliang He, Bo Zhang, Shouzhi Wang
Design and research the sensing terminal system of mass transmission equipment, adopting modular design, including sensing modules. The system has a data analysis module and a protocol analysis module, which are sensed through image acquisition module, positioning module, temperature and humidity detection module and other perception modules. The terminal is connected to the user's mobile terminal through the Bluetooth module, 4G module, and server. The measurement equipment sensing terminal system has functions such as basic file sorting, physical and geographic topology information upload, hierarchical line loss accounting, metering box and power meter power outage report, early warning report of suspected electricity theft, environmental temperature and humidity monitoring, etc.
{"title":"Research on the Computer Aided Design of Sensing Terminal System for Mass Transmission Equipment","authors":"Xin Xia, Chuanliang He, Bo Zhang, Shouzhi Wang","doi":"10.1109/ICESIT53460.2021.9696652","DOIUrl":"https://doi.org/10.1109/ICESIT53460.2021.9696652","url":null,"abstract":"Design and research the sensing terminal system of mass transmission equipment, adopting modular design, including sensing modules. The system has a data analysis module and a protocol analysis module, which are sensed through image acquisition module, positioning module, temperature and humidity detection module and other perception modules. The terminal is connected to the user's mobile terminal through the Bluetooth module, 4G module, and server. The measurement equipment sensing terminal system has functions such as basic file sorting, physical and geographic topology information upload, hierarchical line loss accounting, metering box and power meter power outage report, early warning report of suspected electricity theft, environmental temperature and humidity monitoring, etc.","PeriodicalId":164745,"journal":{"name":"2021 IEEE International Conference on Emergency Science and Information Technology (ICESIT)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114657252","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 : 2021-11-22DOI: 10.1109/ICESIT53460.2021.9696808
Xie Xikun, Liang Changjiang, Xu Meng
Common feature engineering method and traditional machine visual detection algorithm have problems with strong subjective dependence, low detection accuracy and limited detection range in the detection of metal surface defects. Integrated the ECA attention mechanism to realize the adaptive weight assignment in the important areas of the image will form ECAMobileNetV2 as the model backbone feature extraction network, then use the PANet module of YOLOV4 to enhance the defect feature-one lightweight Yolo V 4 model (ECA_MobileNetV2_yoloV4, abb EMV2yoloV4) integrated ECA and MobileNet. Our method got highest detection accuracy, applied the datasets of metal surface defects for defect types in GCT10 and NED_DET, with mAP of 0.86 and 0.68 respectively. it's significantly higher than MV2yoloV4 and MV3yoloV 4 integrating attention mechanism SE. The model parameter reaching 10.4M is less lightweight than novel detection networks such as Efficientdet and Ghost etc. Experexperiment shows that EMV2yolo V 4 better solves the problem of low recognition accuracy caused by background pixels and brightness. The single image inference time of 18.44ms and frame rate up to 54.25f/s. It can meet the requirements of lightweight deployment and accuracy requirements of metal surface defect detection.
{"title":"Application of attention YOLOV 4 algorithm in metal defect detection","authors":"Xie Xikun, Liang Changjiang, Xu Meng","doi":"10.1109/ICESIT53460.2021.9696808","DOIUrl":"https://doi.org/10.1109/ICESIT53460.2021.9696808","url":null,"abstract":"Common feature engineering method and traditional machine visual detection algorithm have problems with strong subjective dependence, low detection accuracy and limited detection range in the detection of metal surface defects. Integrated the ECA attention mechanism to realize the adaptive weight assignment in the important areas of the image will form ECAMobileNetV2 as the model backbone feature extraction network, then use the PANet module of YOLOV4 to enhance the defect feature-one lightweight Yolo V 4 model (ECA_MobileNetV2_yoloV4, abb EMV2yoloV4) integrated ECA and MobileNet. Our method got highest detection accuracy, applied the datasets of metal surface defects for defect types in GCT10 and NED_DET, with mAP of 0.86 and 0.68 respectively. it's significantly higher than MV2yoloV4 and MV3yoloV 4 integrating attention mechanism SE. The model parameter reaching 10.4M is less lightweight than novel detection networks such as Efficientdet and Ghost etc. Experexperiment shows that EMV2yolo V 4 better solves the problem of low recognition accuracy caused by background pixels and brightness. The single image inference time of 18.44ms and frame rate up to 54.25f/s. It can meet the requirements of lightweight deployment and accuracy requirements of metal surface defect detection.","PeriodicalId":164745,"journal":{"name":"2021 IEEE International Conference on Emergency Science and Information Technology (ICESIT)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123052801","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：