Video Anomaly Detection Algorithm Based On Constrained Sparse Representation

Outlier detection,also known as the deviation detection is to detect and discover those abnormal data patterns which are not conformed to normal(expected)data patterns in observed data.According to different applications,these abnormal patterns are noted as outlier,novelty or stain [1].Recently,anomaly detection has been widely used in fault diagnosis,disease detection,identity identification,intelligent monitoring and other fields.At the same time,with the rapid development of the society,security awareness has been continuously strengthened,Meanwhile,image processing,machine vision and other computer and network transmission technology lead to the rapid prosperity of the video surveillance market.Application video surveillance system has been deep into all areas of daily life.Therefore,this article focuses on the video abnormal event detection method,How to intelligently analyze the large amount of high-dimensional video data and effectively find emergency and abnormal events in them not only is the key to improve the intelligence level of the video surveillance system,and has become an important everyday issues relating to people's lives and property security and social stability,It is also one of the most important research topics in the age of big data.In recent decades,video abnormal event detection technology has made great progress,and accumulated a wealth of theory and the emergence of a large number of detection methods.Although the existing video anomaly detection can be obtained superior performance,since the video data is always with the characters of redundant,high-dimensional,higher information capacity and complex relations,fuzzy,strong space-time continuum,fickle and noisy,making the video abnormal events detection becomes a tough task.In this paper,we analyze and summarize the existing sparse representation video abnormal event detection methods.on this basis,we propose a video anomaly detection algorithm based on constrained sparse representation(CSR).The method is mainly bound into the neighborhood graph sparse coding object equation,trained normal video image sparse coding,and then use these coding to detect the video anomaly events.To test the performance of our algorithm,we conducted a comparative test data set on the UMN database.Experimental results demonstrate the effectiveness of our method.And proved the proposed method has better contrast detection.