Xinfeng Zhang, Qiling Ni, Shuhan Chen, Baoqing Yang, Bin Li
The crowd motion in public places is generally disorderly but locally orderly. Therefore, dividing the crowd flow into regions with basically consistent motion states can help us better understand and analyze the crowd's motion states. For this reason, a deep motion transformation network is proposed to segment the crowd flow into different motion states, which avoids the problem of parameter selection based on the clustering method. We test the method in different crowd density scenarios, and the experimental results show that the proposed method can achieve a better segmentation effect than the previous methods.
{"title":"A Crowd Flow Segmentation Method based on Deep Motion Transformation Network","authors":"Xinfeng Zhang, Qiling Ni, Shuhan Chen, Baoqing Yang, Bin Li","doi":"10.1145/3449388.3449396","DOIUrl":"https://doi.org/10.1145/3449388.3449396","url":null,"abstract":"The crowd motion in public places is generally disorderly but locally orderly. Therefore, dividing the crowd flow into regions with basically consistent motion states can help us better understand and analyze the crowd's motion states. For this reason, a deep motion transformation network is proposed to segment the crowd flow into different motion states, which avoids the problem of parameter selection based on the clustering method. We test the method in different crowd density scenarios, and the experimental results show that the proposed method can achieve a better segmentation effect than the previous methods.","PeriodicalId":326682,"journal":{"name":"2021 6th International Conference on Multimedia and Image Processing","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134263396","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}
Texture extraction is considered as a basic but a very challenging work in a lot of computer vision fields. Yet texture is not precisely defined, which is difficult to be separated from edges. In this paper, a novel texture extraction algorithm was proposed. Nonlinear structure tensor was introduced to distinguish textures out from edges. And an 8-neighborhood tensor inhomogeneous average sparse matrix was presented to smooth the images. The smoothness weights are determined by the local anisotropy. By applying this inhomogeneous average sparse matrix to the input images, the textures are smoothed to the detail layer while the edges are remained in the original images. The effectiveness of our method was demonstrated by the comparison results with other existing generally acknowledged texture extraction algorithms. And the sparse matrix framework reduces the computational cost than the convolution frameworks.
{"title":"Tensor Inhomogeneous Average Sparse Matrix Based Texture Extraction","authors":"Xin Jin, Yongxin Jiang, Chengtao Yi","doi":"10.1145/3449388.3449397","DOIUrl":"https://doi.org/10.1145/3449388.3449397","url":null,"abstract":"Texture extraction is considered as a basic but a very challenging work in a lot of computer vision fields. Yet texture is not precisely defined, which is difficult to be separated from edges. In this paper, a novel texture extraction algorithm was proposed. Nonlinear structure tensor was introduced to distinguish textures out from edges. And an 8-neighborhood tensor inhomogeneous average sparse matrix was presented to smooth the images. The smoothness weights are determined by the local anisotropy. By applying this inhomogeneous average sparse matrix to the input images, the textures are smoothed to the detail layer while the edges are remained in the original images. The effectiveness of our method was demonstrated by the comparison results with other existing generally acknowledged texture extraction algorithms. And the sparse matrix framework reduces the computational cost than the convolution frameworks.","PeriodicalId":326682,"journal":{"name":"2021 6th International Conference on Multimedia and Image Processing","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132564753","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}
Nowadays due to the rapid growth of the internet, more and more websites or online applications are created to help people live a more convenient life. The main revenue source for these websites and online applications are through advertisements revenue. It has been a really hot topic recently regarding how to increase advertisements revenue through a better Ads designs that are more attractive to users. This paper researched on a real time Ads performance measure system based on apache flink that can effectively measure Ads performance 15 minutes after the Ads started. We also implemented this system in the paper as well.
{"title":"Research of Advertisement performance measure system based on Apache Flink and AB testing","authors":"Yuelan Liu, Yuefan Liu","doi":"10.1145/3449388.3449402","DOIUrl":"https://doi.org/10.1145/3449388.3449402","url":null,"abstract":"Nowadays due to the rapid growth of the internet, more and more websites or online applications are created to help people live a more convenient life. The main revenue source for these websites and online applications are through advertisements revenue. It has been a really hot topic recently regarding how to increase advertisements revenue through a better Ads designs that are more attractive to users. This paper researched on a real time Ads performance measure system based on apache flink that can effectively measure Ads performance 15 minutes after the Ads started. We also implemented this system in the paper as well.","PeriodicalId":326682,"journal":{"name":"2021 6th International Conference on Multimedia and Image Processing","volume":"108 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124065380","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}
Real-time vehicle detection based traffic monitoring is a hot research topic within the area of computer vision. In view of the problem of low detection accuracy and low processing speed, a vehicle detection method based on Improved Deep Structure is proposed in this study. Due to the characteristics of highway vehicles with a fixed aspect ratio, k-means ++ clustering method is used to select new anchor boxes to eliminate false targets at an early stage followed by improved depth structure with deep sort. Experimental results demonstrated that our proposed method on standard data set KITTI-UA achieved higher precision and faster speed than the existing algorithms.
{"title":"Vehicle Flow Detection Based on Improved Deep Structure and Deep Sort","authors":"Haobin Li, Yi Zhang","doi":"10.1145/3449388.3449394","DOIUrl":"https://doi.org/10.1145/3449388.3449394","url":null,"abstract":"Real-time vehicle detection based traffic monitoring is a hot research topic within the area of computer vision. In view of the problem of low detection accuracy and low processing speed, a vehicle detection method based on Improved Deep Structure is proposed in this study. Due to the characteristics of highway vehicles with a fixed aspect ratio, k-means ++ clustering method is used to select new anchor boxes to eliminate false targets at an early stage followed by improved depth structure with deep sort. Experimental results demonstrated that our proposed method on standard data set KITTI-UA achieved higher precision and faster speed than the existing algorithms.","PeriodicalId":326682,"journal":{"name":"2021 6th International Conference on Multimedia and Image Processing","volume":"136 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116332123","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}
Aiming at the problem of low matching efficiency of traditional AKAZE algorithm, an improved algorithm is proposed that combines AKAZE and FREAK algorithms. First, AKAZE is used to extract feature points to ensure the accuracy of feature detection, and then the FREAK operator is used to calculate the descriptor, and then the VFC algorithm is used to perform accurate matching to improve the matching efficiency, and finally the weighted fusion algorithm is used to fuse the image. The research results show that compared with the traditional SIFT, the improved AKAZE algorithm improves the feature extraction time by about 1.11s, and the improved AKAZE algorithm in terms of computing descriptor efficiency increases the time by 1.32s than the SIFT and AKAZE algorithms, which can get higher The accuracy and matching results of the UAV realize rapid and seamless splicing of UAV images.
{"title":"Research on UAV Image Mosaic Based on Improved AKAZE Feature and VFC Algorithm","authors":"Q. Yan, Qianwen Li, Tongkang Zhang","doi":"10.1145/3449388.3449403","DOIUrl":"https://doi.org/10.1145/3449388.3449403","url":null,"abstract":"Aiming at the problem of low matching efficiency of traditional AKAZE algorithm, an improved algorithm is proposed that combines AKAZE and FREAK algorithms. First, AKAZE is used to extract feature points to ensure the accuracy of feature detection, and then the FREAK operator is used to calculate the descriptor, and then the VFC algorithm is used to perform accurate matching to improve the matching efficiency, and finally the weighted fusion algorithm is used to fuse the image. The research results show that compared with the traditional SIFT, the improved AKAZE algorithm improves the feature extraction time by about 1.11s, and the improved AKAZE algorithm in terms of computing descriptor efficiency increases the time by 1.32s than the SIFT and AKAZE algorithms, which can get higher The accuracy and matching results of the UAV realize rapid and seamless splicing of UAV images.","PeriodicalId":326682,"journal":{"name":"2021 6th International Conference on Multimedia and Image Processing","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122162757","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：