This paper proposes an end-to-end abnormal behavior detection network to detect strenuous movements in slow moving crowds, such as running, bicycling in transportation surveillance videos. The algorithm forms continuous video frames into a video packet and use the video packet feature extractor to obtain the spatio-temporal information. The implicit vector-based attention mechanism will work on the extracted video packet features to highlight the important features. We use fully connected layers to transform the space and reduce the computation. Finally, the packet-pooling maps the processed video packet features to the abnormal scores. The network input is flexible to cope with the form of video streams, and the network output is the abnormal score. The designed compound loss function will help the model improve the classification performance. This paper arranges several commonly used anomaly detection datasets and tests the algorithms on the integrated dataset. The experiment results show that the proposed algorithm has significant advantages in many objective metrics comparing with other anomaly detection algorithms.
{"title":"The anomaly behavior detection algorithm with video-packet attention in transportation surveillance videos","authors":"Liyuan Wang, S. Yu, Ling Ding, Yuanxu Wu, Yu Chen, Jinsheng Xiao","doi":"10.1117/12.2671205","DOIUrl":"https://doi.org/10.1117/12.2671205","url":null,"abstract":"This paper proposes an end-to-end abnormal behavior detection network to detect strenuous movements in slow moving crowds, such as running, bicycling in transportation surveillance videos. The algorithm forms continuous video frames into a video packet and use the video packet feature extractor to obtain the spatio-temporal information. The implicit vector-based attention mechanism will work on the extracted video packet features to highlight the important features. We use fully connected layers to transform the space and reduce the computation. Finally, the packet-pooling maps the processed video packet features to the abnormal scores. The network input is flexible to cope with the form of video streams, and the network output is the abnormal score. The designed compound loss function will help the model improve the classification performance. This paper arranges several commonly used anomaly detection datasets and tests the algorithms on the integrated dataset. The experiment results show that the proposed algorithm has significant advantages in many objective metrics comparing with other anomaly detection algorithms.","PeriodicalId":227528,"journal":{"name":"International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133216732","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}
Global pandemic due to the spread of COVID-19 has post challenges in a new dimension on facial recognition, where people start to wear masks. Under such condition, the authors consider utilizing machine learning in image inpainting to tackle the problem, by complete the possible face that is originally covered in mask. In particular, autoencoder has great potential on retaining important, general features of the image as well as the generative power of the Generative Adversarial Network (GAN). The authors implement a combination of the two models, context encoders and explain how it combines the power of the two models and train the model with 50,000 images of influencers faces and yields a solid result that still contains space for improvements. Furthermore, the authors discuss some shortcomings with the model, their possible improvements, as well as some area of study for future investigation for applicative perspective, as well as directions to further enhance and refine the model.
{"title":"GAN-based algorithm for efficient image inpainting","authors":"Zheng Han, Zehao Jiang, Yuan Ju","doi":"10.1117/12.2671788","DOIUrl":"https://doi.org/10.1117/12.2671788","url":null,"abstract":"Global pandemic due to the spread of COVID-19 has post challenges in a new dimension on facial recognition, where people start to wear masks. Under such condition, the authors consider utilizing machine learning in image inpainting to tackle the problem, by complete the possible face that is originally covered in mask. In particular, autoencoder has great potential on retaining important, general features of the image as well as the generative power of the Generative Adversarial Network (GAN). The authors implement a combination of the two models, context encoders and explain how it combines the power of the two models and train the model with 50,000 images of influencers faces and yields a solid result that still contains space for improvements. Furthermore, the authors discuss some shortcomings with the model, their possible improvements, as well as some area of study for future investigation for applicative perspective, as well as directions to further enhance and refine the model.","PeriodicalId":227528,"journal":{"name":"International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116125058","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}
In the big data environment, the key is the precise recommendation of learning resources to learners. The core is the in-deep mining of learners’ personalized demands. This study solves this problem by constructing learner personas. Primarily, collect web learning data of learners to cluster them. Then analyze the characteristics of learners to predict their learning intentions and knowledge blind spots. Based on it, generate a clear personalized learning path subsequently. Precise positioning, quickly finding out the learner's ability and quality shortcomings. And completing the accurate recommendation to learners. It will help learners establish a reasonable learning path, and provide more accurate service support. This study will provide a theoretical basis for carrying out big data precision services and meeting the personalized learning needs of learners.
{"title":"Research on the construction of learner personas","authors":"Hailan Li, Kongyang Peng, Fengying Shang, Haoli Ren","doi":"10.1117/12.2671043","DOIUrl":"https://doi.org/10.1117/12.2671043","url":null,"abstract":"In the big data environment, the key is the precise recommendation of learning resources to learners. The core is the in-deep mining of learners’ personalized demands. This study solves this problem by constructing learner personas. Primarily, collect web learning data of learners to cluster them. Then analyze the characteristics of learners to predict their learning intentions and knowledge blind spots. Based on it, generate a clear personalized learning path subsequently. Precise positioning, quickly finding out the learner's ability and quality shortcomings. And completing the accurate recommendation to learners. It will help learners establish a reasonable learning path, and provide more accurate service support. This study will provide a theoretical basis for carrying out big data precision services and meeting the personalized learning needs of learners.","PeriodicalId":227528,"journal":{"name":"International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022)","volume":"77 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114863382","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}
Wenteng Liang, Shang Dai, Yizhen You, Kang Yang, Jianan Zhang, Tai Sun, Ruyi Li, Yue Zhang, linxi zou
In order to improve the accuracy of power dispatching text analysis and the ability to guide the operation of the power grid, a power dispatch text entity recognition method is proposed based on Bidirectional Encoder Representations from Transformers-Conditional Random Field (BERT-CRF). Taking the power grid fault handling plan text as the research object, the entity marking method of the fault handling plan is proposed. The word vector of the plan entity is calculated based on the BERT pre-training model, the characterization ability of the professional entity of the plan is enhanced by fine-tuning the initial BERT parameters, and the recognition ability of the plan text sequence is improved from the overall situation to access the CRF layer in the neural network. Thus, an entity recognition model of fault handling plan is established based on the BERT-CRF. Through the verification of a power grid fault handling plan, the proposed method has higher power dispatch entity and event recognition accuracy compared with other algorithms.
{"title":"The entity and event recognition method of power dispatching text information based on BERT-CRF","authors":"Wenteng Liang, Shang Dai, Yizhen You, Kang Yang, Jianan Zhang, Tai Sun, Ruyi Li, Yue Zhang, linxi zou","doi":"10.1117/12.2671453","DOIUrl":"https://doi.org/10.1117/12.2671453","url":null,"abstract":"In order to improve the accuracy of power dispatching text analysis and the ability to guide the operation of the power grid, a power dispatch text entity recognition method is proposed based on Bidirectional Encoder Representations from Transformers-Conditional Random Field (BERT-CRF). Taking the power grid fault handling plan text as the research object, the entity marking method of the fault handling plan is proposed. The word vector of the plan entity is calculated based on the BERT pre-training model, the characterization ability of the professional entity of the plan is enhanced by fine-tuning the initial BERT parameters, and the recognition ability of the plan text sequence is improved from the overall situation to access the CRF layer in the neural network. Thus, an entity recognition model of fault handling plan is established based on the BERT-CRF. Through the verification of a power grid fault handling plan, the proposed method has higher power dispatch entity and event recognition accuracy compared with other algorithms.","PeriodicalId":227528,"journal":{"name":"International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116046658","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}
For embedded modern equipment, the current gait recognition algorithm model is difficult to deploy on it due to a large amount of gait frame image data, slow network processing speed, complex structure and low computational efficiency. In this paper, a lightweight convolutional network model integrating the attention mechanism is proposed. The algorithm first performs morphological processing on the image, extracts the gait contour image, and calculates the gait energy image; integrates the attention mechanism with MobileNetV1. The feature information of the image is effectively extracted, and the parameters of the network are reduced. A number of body method validation experiments are conducted in the CAISIA-B gait database of the Chinese Academy of Sciences, and the experimental results are significantly improved with other deep learning models.
{"title":"Human gait recognition algorithm based on MobileNetV1 with attention mechanism","authors":"Jinsha Zhang, Xuedong Zhang","doi":"10.1117/12.2671349","DOIUrl":"https://doi.org/10.1117/12.2671349","url":null,"abstract":"For embedded modern equipment, the current gait recognition algorithm model is difficult to deploy on it due to a large amount of gait frame image data, slow network processing speed, complex structure and low computational efficiency. In this paper, a lightweight convolutional network model integrating the attention mechanism is proposed. The algorithm first performs morphological processing on the image, extracts the gait contour image, and calculates the gait energy image; integrates the attention mechanism with MobileNetV1. The feature information of the image is effectively extracted, and the parameters of the network are reduced. A number of body method validation experiments are conducted in the CAISIA-B gait database of the Chinese Academy of Sciences, and the experimental results are significantly improved with other deep learning models.","PeriodicalId":227528,"journal":{"name":"International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117081564","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}
Diversity among the members of classifiers is deemed to be a key point in classifier ensemble. However, there doesn’t exist a widely accepted diversity measure and construct. In this paper, we propose a sample and feature double random construction of training sample variability. A support vector machine is used as the base classifier to construct the difference by distinguishing the regularization term C and the kernel function. Based on the negative correlation theory, the base classifier generalization error and disparity judgment functions are proposed, and the base classifier is integrated by ranking according to the judgment functions, which could achieve a higher accuracy rate by the support vector machine ensemble.
{"title":"Evaluate the performance of the support vector machines ensemble","authors":"Bowen Liu, Yihui Qiu","doi":"10.1117/12.2671159","DOIUrl":"https://doi.org/10.1117/12.2671159","url":null,"abstract":"Diversity among the members of classifiers is deemed to be a key point in classifier ensemble. However, there doesn’t exist a widely accepted diversity measure and construct. In this paper, we propose a sample and feature double random construction of training sample variability. A support vector machine is used as the base classifier to construct the difference by distinguishing the regularization term C and the kernel function. Based on the negative correlation theory, the base classifier generalization error and disparity judgment functions are proposed, and the base classifier is integrated by ranking according to the judgment functions, which could achieve a higher accuracy rate by the support vector machine ensemble.","PeriodicalId":227528,"journal":{"name":"International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022)","volume":"46 4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116312198","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}
Expressway project is usually built in extremely complex natural and cultural environment. The whole process of project implementation management is a continuous and dynamic management practice process, which will be affected by internal and external uncertainties, and may directly affect the benefit and even the survival and development of enterprises. Therefore, this paper studies and analyzes the risk of investment in the highway project and several factors that may affect it. This paper selects the actual situation of 112 expressways in China and analyzes them through 30 different risk indexes. Through constructing multiple linear regression model, the factors that may affect the investment risk of expressway project are analyzed. Finally, there are 20 risk indicators to influence the investment risk of expressway project, and this paper constructs the weight model of expressway investment risk evaluation hierarchy and tries to verify it.
{"title":"Analysis of influencing factors on investment risk of expressway project in China","authors":"Liangjie Wu, Yangyang Li, lianlian shang","doi":"10.1117/12.2671195","DOIUrl":"https://doi.org/10.1117/12.2671195","url":null,"abstract":"Expressway project is usually built in extremely complex natural and cultural environment. The whole process of project implementation management is a continuous and dynamic management practice process, which will be affected by internal and external uncertainties, and may directly affect the benefit and even the survival and development of enterprises. Therefore, this paper studies and analyzes the risk of investment in the highway project and several factors that may affect it. This paper selects the actual situation of 112 expressways in China and analyzes them through 30 different risk indexes. Through constructing multiple linear regression model, the factors that may affect the investment risk of expressway project are analyzed. Finally, there are 20 risk indicators to influence the investment risk of expressway project, and this paper constructs the weight model of expressway investment risk evaluation hierarchy and tries to verify it.","PeriodicalId":227528,"journal":{"name":"International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022)","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115069218","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}
Due to the limitation of computer capacity and energy of equipment, unmanned equipment cannot perform intensive computer tasks well during emergency failure inspection. In order to solve the above problems, this paper proposes a task waste strategy based on Deep Reinforcement Learning (DRL), which is mainly applicable to several UAVs and individual ES scenarios. First of all, an end edge cloud cooperative unloading architecture is built in the edge environment of UAV, and the problem of unloading tasks is classified as an optimization problem to achieve the minimum delay under the limit of the computing and communication resources of the Edge Server (ES). Secondly, the problem is constructed as Markov decision, and Deep Q Network (DQN) is used to solve the optimization problem, and experience playback mechanism and greedy algorithm are introduced into the learning process. Experiments show that the mitigation strategy has lower latency and higher reliability.
{"title":"Offloading strategy for UAV power inspection task based on deep reinforcement learning","authors":"Tong Jin, Gu Minghao, Sha Yun, Deng Fang-ming","doi":"10.1117/12.2671522","DOIUrl":"https://doi.org/10.1117/12.2671522","url":null,"abstract":"Due to the limitation of computer capacity and energy of equipment, unmanned equipment cannot perform intensive computer tasks well during emergency failure inspection. In order to solve the above problems, this paper proposes a task waste strategy based on Deep Reinforcement Learning (DRL), which is mainly applicable to several UAVs and individual ES scenarios. First of all, an end edge cloud cooperative unloading architecture is built in the edge environment of UAV, and the problem of unloading tasks is classified as an optimization problem to achieve the minimum delay under the limit of the computing and communication resources of the Edge Server (ES). Secondly, the problem is constructed as Markov decision, and Deep Q Network (DQN) is used to solve the optimization problem, and experience playback mechanism and greedy algorithm are introduced into the learning process. Experiments show that the mitigation strategy has lower latency and higher reliability.","PeriodicalId":227528,"journal":{"name":"International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022)","volume":"13 5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115489426","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}
Lithium-ion battery has become an indispensable energy storage component in our life because of its environmental protection and high energy characteristics. The battery SOH is the decisive factor to ensure its stability. For the sake to improve the accuracy of EV battery SOH prediction. Firstly, data structuring, PCA dimension reduction and data standardization were used to transform downloaded data into data that could be trained with high accuracy model. After that, the characteristic factors related to battery capacity were extracted from the battery charging data and correlation analysis was carried out. According to the method of Pearson coefficient, the features with strong correlation were left and then imported into the sample data. The factor parameters of SVR and other models were optimized by grid search algorithm, and the final prediction model was established. Lithium-ion battery has become an indispensable energy storage component in our life because of its environmental protection and high energy characteristics. The battery SOH is the decisive factor to ensure its stability.
{"title":"Battery health analysis of electric vehicle based on EL-SVR","authors":"Ling Zhong, X. Liu","doi":"10.1117/12.2671142","DOIUrl":"https://doi.org/10.1117/12.2671142","url":null,"abstract":"Lithium-ion battery has become an indispensable energy storage component in our life because of its environmental protection and high energy characteristics. The battery SOH is the decisive factor to ensure its stability. For the sake to improve the accuracy of EV battery SOH prediction. Firstly, data structuring, PCA dimension reduction and data standardization were used to transform downloaded data into data that could be trained with high accuracy model. After that, the characteristic factors related to battery capacity were extracted from the battery charging data and correlation analysis was carried out. According to the method of Pearson coefficient, the features with strong correlation were left and then imported into the sample data. The factor parameters of SVR and other models were optimized by grid search algorithm, and the final prediction model was established. Lithium-ion battery has become an indispensable energy storage component in our life because of its environmental protection and high energy characteristics. The battery SOH is the decisive factor to ensure its stability.","PeriodicalId":227528,"journal":{"name":"International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022)","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125016157","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}
Face recognition has been widely used in daily life, but the existing model systems use processed high-quality datasets in training, while the face pictures in real scenes usually contain the influence of blurring, lighting, obscuring and other factors, thus making the existing face recognition models cannot perform well, and secondly, the existing face datasets have less data of Asian descent, resulting in the distribution learned by the models with the actual application. There is a certain error in the actual application. We propose a method to train face recognition models for realistic scenes by image augment of local face data to improve the classification accuracy of the models for low-quality images, and we demonstrate the feasibility of our method through experiments. Our method improves 0.619% and 0.414% in classifying images with added illumination and added random squares, respectively, compared to the current state-of-the-art methods.
{"title":"A training method for face representation models in realistic scenarios","authors":"C. Li","doi":"10.1117/12.2671250","DOIUrl":"https://doi.org/10.1117/12.2671250","url":null,"abstract":"Face recognition has been widely used in daily life, but the existing model systems use processed high-quality datasets in training, while the face pictures in real scenes usually contain the influence of blurring, lighting, obscuring and other factors, thus making the existing face recognition models cannot perform well, and secondly, the existing face datasets have less data of Asian descent, resulting in the distribution learned by the models with the actual application. There is a certain error in the actual application. We propose a method to train face recognition models for realistic scenes by image augment of local face data to improve the classification accuracy of the models for low-quality images, and we demonstrate the feasibility of our method through experiments. Our method improves 0.619% and 0.414% in classifying images with added illumination and added random squares, respectively, compared to the current state-of-the-art methods.","PeriodicalId":227528,"journal":{"name":"International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022)","volume":"49 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121467372","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：