{"title":"Multi-object Foreground Extraction in Streaming Video using Low Rank Sparse Decomposition","authors":"Yogesh Sanku, Soumyo Bhattacharjee, Saumik Bhattacharya","doi":"10.1109/INDICON52576.2021.9691672","DOIUrl":null,"url":null,"abstract":"Low rank sparse decomposition (LRSD) algorithm is a popular technique to split an input video in to a low rank form and a complementary sparse form. The decomposed low rank matrix signifies the background information while the sparse matrix captures the foreground information. The real power of the algorithm proposed is in the use of stationary camera systems, particularly in surveillance systems to extract moving objects efficiently for analyses. However, the existing LRSD algorithms are designed such that it can only work on the entire video cube, but not on streaming videos. This severely affects the usability of LRSD-based algorithms in real-world surveillance tasks. In this paper, we propose a novel LRSD decomposition algorithm that can deal with streaming video data. To the best of our knowledge, this is the first attempt to design an LRSD-based system to work on streaming videos with varying background conditions. Exhaustive experimental analyses have shown that the proposed framework can process the videos almost in real-time.","PeriodicalId":106004,"journal":{"name":"2021 IEEE 18th India Council International Conference (INDICON)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2021-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 IEEE 18th India Council International Conference (INDICON)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/INDICON52576.2021.9691672","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 1
Abstract
Low rank sparse decomposition (LRSD) algorithm is a popular technique to split an input video in to a low rank form and a complementary sparse form. The decomposed low rank matrix signifies the background information while the sparse matrix captures the foreground information. The real power of the algorithm proposed is in the use of stationary camera systems, particularly in surveillance systems to extract moving objects efficiently for analyses. However, the existing LRSD algorithms are designed such that it can only work on the entire video cube, but not on streaming videos. This severely affects the usability of LRSD-based algorithms in real-world surveillance tasks. In this paper, we propose a novel LRSD decomposition algorithm that can deal with streaming video data. To the best of our knowledge, this is the first attempt to design an LRSD-based system to work on streaming videos with varying background conditions. Exhaustive experimental analyses have shown that the proposed framework can process the videos almost in real-time.