Pub Date : 2022-05-20DOI: 10.1109/cvidliccea56201.2022.9824838
Jingrong Wang, Limeng Lu, Zixiang Zhang, Nady Slam
Since the outbreak of novel coronavirus pneumonia (COVID-19) in 2019, normal learning and living have been severely affected, and human life and health have been seriously threatened. Therefore, it is crucial to diagnose the novel coronavirus pneumonia rapidly and efficiently. In this study, based on the classical image classification neural network model, a novel deep convolutional neural network model based on the attention mechanism is proposed and named the LACNN_CBAM model. The accuracy Acc, precision Pre, recall Rec and F-1 scores of the model in the public dataset collated from published papers are 0.989, 0.992, 0.992, and 0.992, which are respectively higher than existing learning models. The model determines whether a patient has COVID-19 and community-acquired pneumonia by patient’s CT images. The effectiveness of the model was demonstrated by experimental results on a clinical dataset. We believe that the model proposed in this paper can help physicians to diagnose COVID-19 and community-acquired pneumonia efficiently and accurately in reality.
{"title":"A Novel Deep Convolution Neural Network Model for CT Image Classification Based on COVID-19","authors":"Jingrong Wang, Limeng Lu, Zixiang Zhang, Nady Slam","doi":"10.1109/cvidliccea56201.2022.9824838","DOIUrl":"https://doi.org/10.1109/cvidliccea56201.2022.9824838","url":null,"abstract":"Since the outbreak of novel coronavirus pneumonia (COVID-19) in 2019, normal learning and living have been severely affected, and human life and health have been seriously threatened. Therefore, it is crucial to diagnose the novel coronavirus pneumonia rapidly and efficiently. In this study, based on the classical image classification neural network model, a novel deep convolutional neural network model based on the attention mechanism is proposed and named the LACNN_CBAM model. The accuracy Acc, precision Pre, recall Rec and F-1 scores of the model in the public dataset collated from published papers are 0.989, 0.992, 0.992, and 0.992, which are respectively higher than existing learning models. The model determines whether a patient has COVID-19 and community-acquired pneumonia by patient’s CT images. The effectiveness of the model was demonstrated by experimental results on a clinical dataset. We believe that the model proposed in this paper can help physicians to diagnose COVID-19 and community-acquired pneumonia efficiently and accurately in reality.","PeriodicalId":23649,"journal":{"name":"Vision","volume":"105 1","pages":"15-20"},"PeriodicalIF":0.0,"publicationDate":"2022-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88989723","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 : 2022-05-20DOI: 10.1109/cvidliccea56201.2022.9824528
Yuechen Hao
With the explosion of Nakamoto’s paper, blockchain technology has developed rapidly, but at the same time, security problems are emerging one after another. As a potential security hazard in the payment field, 51% attack brings huge risks to the normal operation of the blockchain system. Miners with great computing power have the ability to monopolize the generation of blocks and modify the generated blocks. Therefore, it is necessary to do research of this kind of attacks. This article cites both Nakamoto model and Rosenfeld model to illustrate relationship between computing power and attack success rate. Through a series of mining experiments, this paper preliminarily introduces the operation principles of blockchain based on Ethereum, including proof of stake, smart contracts etc. Models show that for rational attackers who pursue interests, they lack the motivation to launch 51% attacks. For attackers who only destroy the bitcoin system, they need to master huge financial resources to launch 51% of attacks, and even need financial support at the national level, which is very difficult. It can be said that 51% attacks against bitcoin are only theoretically possible, but users still need to pay enough attention. As an emerging technology, blockchain technology is currently in the research and exploration stage. While it is applied in the financial field, it is also expanding to other fields. In the future, blockchain technology will not only be used to solve the trust and security problems in the centralized service architecture, but also appear in more decentralized service scenarios. So, the research on blockchain security is particularly important.
{"title":"Research of the 51% attack based on blockchain","authors":"Yuechen Hao","doi":"10.1109/cvidliccea56201.2022.9824528","DOIUrl":"https://doi.org/10.1109/cvidliccea56201.2022.9824528","url":null,"abstract":"With the explosion of Nakamoto’s paper, blockchain technology has developed rapidly, but at the same time, security problems are emerging one after another. As a potential security hazard in the payment field, 51% attack brings huge risks to the normal operation of the blockchain system. Miners with great computing power have the ability to monopolize the generation of blocks and modify the generated blocks. Therefore, it is necessary to do research of this kind of attacks. This article cites both Nakamoto model and Rosenfeld model to illustrate relationship between computing power and attack success rate. Through a series of mining experiments, this paper preliminarily introduces the operation principles of blockchain based on Ethereum, including proof of stake, smart contracts etc. Models show that for rational attackers who pursue interests, they lack the motivation to launch 51% attacks. For attackers who only destroy the bitcoin system, they need to master huge financial resources to launch 51% of attacks, and even need financial support at the national level, which is very difficult. It can be said that 51% attacks against bitcoin are only theoretically possible, but users still need to pay enough attention. As an emerging technology, blockchain technology is currently in the research and exploration stage. While it is applied in the financial field, it is also expanding to other fields. In the future, blockchain technology will not only be used to solve the trust and security problems in the centralized service architecture, but also appear in more decentralized service scenarios. So, the research on blockchain security is particularly important.","PeriodicalId":23649,"journal":{"name":"Vision","volume":"09 1","pages":"278-283"},"PeriodicalIF":0.0,"publicationDate":"2022-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86502960","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 : 2022-05-20DOI: 10.1109/cvidliccea56201.2022.9825072
Pnegyu Dai
Vision-based solutions for target detection in autonomous driving are very much about the accuracy of detection. A correct or incorrect detection may cause or avoid a traffic accident. Therefore, in this paper, to further improve the detection accuracy of vision schemes, we propose a multi-model fusion network: Fusion Network with YoloV5 and CBNEet Faster-RCNN (FYCFNet) that fuses a one-stage target detection model and a two-stage model, which consists of three parts: the first part is a single-stage YOLOV5 [1] detection model, the second part is a Faster-RCNN [2] with CBNet-V2 [3] as the backbone, and the third part is the post-fusion head of weighted boxes fusion. We tested the performance of this network and compared it with other mainstream networks, and verified that the network achieves a very impressive accuracy improvement.
{"title":"FYCFNet: Vehicle and Pedestrian Detection Network based on Multi-model Fusion","authors":"Pnegyu Dai","doi":"10.1109/cvidliccea56201.2022.9825072","DOIUrl":"https://doi.org/10.1109/cvidliccea56201.2022.9825072","url":null,"abstract":"Vision-based solutions for target detection in autonomous driving are very much about the accuracy of detection. A correct or incorrect detection may cause or avoid a traffic accident. Therefore, in this paper, to further improve the detection accuracy of vision schemes, we propose a multi-model fusion network: Fusion Network with YoloV5 and CBNEet Faster-RCNN (FYCFNet) that fuses a one-stage target detection model and a two-stage model, which consists of three parts: the first part is a single-stage YOLOV5 [1] detection model, the second part is a Faster-RCNN [2] with CBNet-V2 [3] as the backbone, and the third part is the post-fusion head of weighted boxes fusion. We tested the performance of this network and compared it with other mainstream networks, and verified that the network achieves a very impressive accuracy improvement.","PeriodicalId":23649,"journal":{"name":"Vision","volume":"31 1","pages":"230-236"},"PeriodicalIF":0.0,"publicationDate":"2022-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84995626","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 : 2022-05-20DOI: 10.1109/cvidliccea56201.2022.9825193
Ming Xin, Wenjie Sun, Kaifang Li, Guancheng Hui
Resistance to object appearance deformation and local occlusion is still one of the challenges of multi-object tracking algorithms. Most popular algorithms rely on time-consuming numerical optimization and complex manual design strategies to integrate object appearance information and motion information, so as to alleviate the adverse effects of object appearance deformation and local occlusion on the trajectory updating. This paper proposes a Spatial-Temporal Correlation Memory (STCM) module which can adaptively aggregate useful information from rich historical information in memory. By mining the time dimension information, the STCM module can guide the backbone network to extract the current frame effectively, and adapt to the change in the object’s appearance in the tracking process. Specifically, the STCM module can record the foreground-background information in the history frames and direct the backbone network to focus on the useful information in the current frame. Experiments on the MOT17 data set show that our method outperforms the baseline method and current advanced method in index MOTA and IDFI.
{"title":"Multi-Object Tracking with Spatial-Temporal Correlation Memory Networks","authors":"Ming Xin, Wenjie Sun, Kaifang Li, Guancheng Hui","doi":"10.1109/cvidliccea56201.2022.9825193","DOIUrl":"https://doi.org/10.1109/cvidliccea56201.2022.9825193","url":null,"abstract":"Resistance to object appearance deformation and local occlusion is still one of the challenges of multi-object tracking algorithms. Most popular algorithms rely on time-consuming numerical optimization and complex manual design strategies to integrate object appearance information and motion information, so as to alleviate the adverse effects of object appearance deformation and local occlusion on the trajectory updating. This paper proposes a Spatial-Temporal Correlation Memory (STCM) module which can adaptively aggregate useful information from rich historical information in memory. By mining the time dimension information, the STCM module can guide the backbone network to extract the current frame effectively, and adapt to the change in the object’s appearance in the tracking process. Specifically, the STCM module can record the foreground-background information in the history frames and direct the backbone network to focus on the useful information in the current frame. Experiments on the MOT17 data set show that our method outperforms the baseline method and current advanced method in index MOTA and IDFI.","PeriodicalId":23649,"journal":{"name":"Vision","volume":"32 1","pages":"616-619"},"PeriodicalIF":0.0,"publicationDate":"2022-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89327953","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}
Convolutional neural network has powerful feature learning capabilities and are widely used in the field of image classification. In this paper, an image classification method with improved CNN activation function is proposed. By analyzing the shallow convolutional neural network, a CIFAR-10 image classification model is constructed. In the process of data preprocessing, the digital standardization of the images is completed and the sample labels are one-hot encoded. The model network structure proposed in this paper adopts the ReLU nonlinear activation function and maximum pooling. The training results show the accuracy of the classification model is significantly improved. At the end of this paper, the accuracy rates of the four activation functions of Sigmoid, Tanh, ReLU, and T-ReLU are compared, and the advantages of the unsaturated nonlinear activation function are pointed out. The model is improved by using the T-ReLU activation function, with the accuracy rate increasing from 62% to 76.52%.
{"title":"Design and implementation of improved CNN activation function","authors":"Yihang Tang, Lu Tian, Yichen Liu, YuJieEr Wen, Keyi Kang, Xiyan Zhao","doi":"10.1109/cvidliccea56201.2022.9824061","DOIUrl":"https://doi.org/10.1109/cvidliccea56201.2022.9824061","url":null,"abstract":"Convolutional neural network has powerful feature learning capabilities and are widely used in the field of image classification. In this paper, an image classification method with improved CNN activation function is proposed. By analyzing the shallow convolutional neural network, a CIFAR-10 image classification model is constructed. In the process of data preprocessing, the digital standardization of the images is completed and the sample labels are one-hot encoded. The model network structure proposed in this paper adopts the ReLU nonlinear activation function and maximum pooling. The training results show the accuracy of the classification model is significantly improved. At the end of this paper, the accuracy rates of the four activation functions of Sigmoid, Tanh, ReLU, and T-ReLU are compared, and the advantages of the unsaturated nonlinear activation function are pointed out. The model is improved by using the T-ReLU activation function, with the accuracy rate increasing from 62% to 76.52%.","PeriodicalId":23649,"journal":{"name":"Vision","volume":"33 1","pages":"1166-1170"},"PeriodicalIF":0.0,"publicationDate":"2022-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84267952","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 : 2022-05-20DOI: 10.1109/cvidliccea56201.2022.9825157
Y. Lin
The packet scheduling problem is a classical multidimensional decision problem that requires rational decisions on the inbound as well as the outbound timing of a huge number of packets. With the advent of programmable scheduling, the deployment of packet scheduling algorithms on switches is easier, and multiple scheduling algorithms can be implemented without changing the hardware architecture. The advent of programmable scheduling simplifies the testing and deployment of new scheduling algorithms and can make the application of packet scheduling algorithms much easier to implement.
{"title":"Research on the Development of Programable Packet Scheduling","authors":"Y. Lin","doi":"10.1109/cvidliccea56201.2022.9825157","DOIUrl":"https://doi.org/10.1109/cvidliccea56201.2022.9825157","url":null,"abstract":"The packet scheduling problem is a classical multidimensional decision problem that requires rational decisions on the inbound as well as the outbound timing of a huge number of packets. With the advent of programmable scheduling, the deployment of packet scheduling algorithms on switches is easier, and multiple scheduling algorithms can be implemented without changing the hardware architecture. The advent of programmable scheduling simplifies the testing and deployment of new scheduling algorithms and can make the application of packet scheduling algorithms much easier to implement.","PeriodicalId":23649,"journal":{"name":"Vision","volume":"16 1","pages":"1098-1101"},"PeriodicalIF":0.0,"publicationDate":"2022-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84333132","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 : 2022-05-20DOI: 10.1109/cvidliccea56201.2022.9825235
Pengfei Ji, Dandan Song
Cross-domain sentiment analysis (CDSA) is an essential subtask of sentiment analysis. It aims to utilize rich source domain data to conquer the data-hungry problem on target domain. Most existing approaches depending on deep learning mainly concentrate on common features or pivots. However, few of them consider the effect of external Knowledge Graph (KG). In this paper, we propose a Dual Knowledge Aggregation Network for Cross-Domain Sentiment Analysis (DKAN), which leverages prior knowledge from two external KGs. Specifically, DKAN comprises two main parts. One is extracting sentence representation features. The other aims to introduce external knowledge better. Also, we use SenticNet to avoid noise from KG by selecting top-n words and inserting special tokens in sentences. We also conduct empirical analyses on the effectiveness of our model on the Amazon reviews dataset. DKAN achieves promising performance compared with other methods.
{"title":"A Dual Knowledge Aggregation Network for Cross-Domain Sentiment Analysis","authors":"Pengfei Ji, Dandan Song","doi":"10.1109/cvidliccea56201.2022.9825235","DOIUrl":"https://doi.org/10.1109/cvidliccea56201.2022.9825235","url":null,"abstract":"Cross-domain sentiment analysis (CDSA) is an essential subtask of sentiment analysis. It aims to utilize rich source domain data to conquer the data-hungry problem on target domain. Most existing approaches depending on deep learning mainly concentrate on common features or pivots. However, few of them consider the effect of external Knowledge Graph (KG). In this paper, we propose a Dual Knowledge Aggregation Network for Cross-Domain Sentiment Analysis (DKAN), which leverages prior knowledge from two external KGs. Specifically, DKAN comprises two main parts. One is extracting sentence representation features. The other aims to introduce external knowledge better. Also, we use SenticNet to avoid noise from KG by selecting top-n words and inserting special tokens in sentences. We also conduct empirical analyses on the effectiveness of our model on the Amazon reviews dataset. DKAN achieves promising performance compared with other methods.","PeriodicalId":23649,"journal":{"name":"Vision","volume":"1995 1","pages":"1-7"},"PeriodicalIF":0.0,"publicationDate":"2022-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88106068","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 : 2022-05-20DOI: 10.1109/cvidliccea56201.2022.9824776
Haijia Chen, Dongliang Guan
Shadow removal and restoration of the image content in the shadowed regions by using shadow removal has become more and more popular in computer vision. But almost all current shadow-removal approaches use shadow-free images for training. Recently, an innovative approach that trains samples without this requirement due to this method crops patches with and without shadows from shadow images. However, it is insufficient to directly learn the essential relationships between shadow and shadow-free domains using adversarial learning and cycle-consistency constraints. Moreover, constructing many of these unpaired patches is still time-consuming and laborious. In our paper, we propose a new method named 2Cycles-G2R-ShadowNet. A shadow mask is used in our framework. We use the mask to guide the shadow generation to reformulate cycle-consistency constraints. To weakly-supervised shadow removal, we train shadow images and corresponding masks to leverage shadow generation. In our 2Cycles-G2R-ShadowNet, three subnetworks are used for shadow generation, shadow removal, and image post-processing, and we jointly train and test them end-to-end. Our method can optimize the performance by simultaneously learning to produce shadow masks and remove shadows. Extensive experiments on the ISTD dataset show that 2Cycles-G2R-ShadowNet achieves competitive performances and outperforms the current state of arts and patch-based shadow-removal method.
{"title":"Shadow Removal Based on 2Cycles-GAN","authors":"Haijia Chen, Dongliang Guan","doi":"10.1109/cvidliccea56201.2022.9824776","DOIUrl":"https://doi.org/10.1109/cvidliccea56201.2022.9824776","url":null,"abstract":"Shadow removal and restoration of the image content in the shadowed regions by using shadow removal has become more and more popular in computer vision. But almost all current shadow-removal approaches use shadow-free images for training. Recently, an innovative approach that trains samples without this requirement due to this method crops patches with and without shadows from shadow images. However, it is insufficient to directly learn the essential relationships between shadow and shadow-free domains using adversarial learning and cycle-consistency constraints. Moreover, constructing many of these unpaired patches is still time-consuming and laborious. In our paper, we propose a new method named 2Cycles-G2R-ShadowNet. A shadow mask is used in our framework. We use the mask to guide the shadow generation to reformulate cycle-consistency constraints. To weakly-supervised shadow removal, we train shadow images and corresponding masks to leverage shadow generation. In our 2Cycles-G2R-ShadowNet, three subnetworks are used for shadow generation, shadow removal, and image post-processing, and we jointly train and test them end-to-end. Our method can optimize the performance by simultaneously learning to produce shadow masks and remove shadows. Extensive experiments on the ISTD dataset show that 2Cycles-G2R-ShadowNet achieves competitive performances and outperforms the current state of arts and patch-based shadow-removal method.","PeriodicalId":23649,"journal":{"name":"Vision","volume":"39 1","pages":"552-558"},"PeriodicalIF":0.0,"publicationDate":"2022-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88353242","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 : 2022-05-20DOI: 10.1109/cvidliccea56201.2022.9824731
Teng Wang, Xinyu Liu, Songming Guo, Baishuo Han, Wenhui Yang
With the arrival of digital agriculture,tens of billions agricultural IoT devices are connected to the IoT, and the accompanying problems such as data information of the whole industrial chain of products being vulnerable to tampering. In this paper, based on the problems of low efficiency and poor security of traditional traceability systems, we design an agricultural product traceability framework based on blockchain and IoT.And complete the deployment of blockchain under the open source distributed leger Hyperledger Fabric.Finally, we carry out web application development through Springboot framework to realize the agricultural product traceability system. In addition,we also conducted efficiency as well as performance analysis in the Caliper performance testing framework, and the results showed that the system improved the efficiency of agricultural product information transmission and data security, and had a significant effect on solving the problems of low efficiency and poor security of traditional traceability systems, meeting the needs of practical applications.
{"title":"Blockchain and IoT based traceability system for agricultural products","authors":"Teng Wang, Xinyu Liu, Songming Guo, Baishuo Han, Wenhui Yang","doi":"10.1109/cvidliccea56201.2022.9824731","DOIUrl":"https://doi.org/10.1109/cvidliccea56201.2022.9824731","url":null,"abstract":"With the arrival of digital agriculture,tens of billions agricultural IoT devices are connected to the IoT, and the accompanying problems such as data information of the whole industrial chain of products being vulnerable to tampering. In this paper, based on the problems of low efficiency and poor security of traditional traceability systems, we design an agricultural product traceability framework based on blockchain and IoT.And complete the deployment of blockchain under the open source distributed leger Hyperledger Fabric.Finally, we carry out web application development through Springboot framework to realize the agricultural product traceability system. In addition,we also conducted efficiency as well as performance analysis in the Caliper performance testing framework, and the results showed that the system improved the efficiency of agricultural product information transmission and data security, and had a significant effect on solving the problems of low efficiency and poor security of traditional traceability systems, meeting the needs of practical applications.","PeriodicalId":23649,"journal":{"name":"Vision","volume":"1 1","pages":"316-321"},"PeriodicalIF":0.0,"publicationDate":"2022-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86634784","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 : 2022-05-20DOI: 10.1109/cvidliccea56201.2022.9825352
ShiQi Xi, Chenjie Su, Xiaodong Cheng, Xi Li
The individual identification of dairy cows is of great significance to the development of modern intelligent animal husbandry. It is of great help in remotely monitoring the individual health status of dairy cows and promoting the field of live dairy cattle leasing. Traditional methods of individual identification of dairy cows rely on manual identification, or artificial feature extraction of cow activity data so the accuracy of individual identification of dairy cows cannot be guaranteed. Aiming at this problem, this paper proposes a classification method based on Gramian Angle Field and Migrating Convolutional Neural Networks. By transforming the activity data of 20 cows for 56 days into the Gramian Angle Field and converting it into a three-dimensional image, the time dependence and correlation of the cow activity data are preserved. Combined with the idea of migration learning, a model called MCNN based on VGG16 is proposed. The MCNN model of the generated cow images is classified. The experimental results show that the classification accuracy of this method is about 99.3%, and the classification time is short, which can effectively realize the individual identification of dairy cows.
{"title":"Individual identification of dairy cows based on Gramian Angular Field and Migrating Convolutional Neural Networks","authors":"ShiQi Xi, Chenjie Su, Xiaodong Cheng, Xi Li","doi":"10.1109/cvidliccea56201.2022.9825352","DOIUrl":"https://doi.org/10.1109/cvidliccea56201.2022.9825352","url":null,"abstract":"The individual identification of dairy cows is of great significance to the development of modern intelligent animal husbandry. It is of great help in remotely monitoring the individual health status of dairy cows and promoting the field of live dairy cattle leasing. Traditional methods of individual identification of dairy cows rely on manual identification, or artificial feature extraction of cow activity data so the accuracy of individual identification of dairy cows cannot be guaranteed. Aiming at this problem, this paper proposes a classification method based on Gramian Angle Field and Migrating Convolutional Neural Networks. By transforming the activity data of 20 cows for 56 days into the Gramian Angle Field and converting it into a three-dimensional image, the time dependence and correlation of the cow activity data are preserved. Combined with the idea of migration learning, a model called MCNN based on VGG16 is proposed. The MCNN model of the generated cow images is classified. The experimental results show that the classification accuracy of this method is about 99.3%, and the classification time is short, which can effectively realize the individual identification of dairy cows.","PeriodicalId":23649,"journal":{"name":"Vision","volume":"2015 1","pages":"135-139"},"PeriodicalIF":0.0,"publicationDate":"2022-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86866015","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：