Research And System Realization Of Abnormal Events Detection In Intelligence Video Surveillance

Abnormal event detection is one of the important tasks in computer vision.Its goal is to identify and locate subsequences containing abnormal events in a video sequence.With the continuous increase of monitoring equipment in daily life and industry,the requirements for the intelligence of the monitoring system are also increasing.As one of the main functions of the monitoring system,abnormal event detection is getting more and more attention.However,abnormal event detection is currently facing many challenges,not only the lack of open source tools but also the insufficient use of event features by the algorithm itself.Therefore,in response to the current challenges,this paper designs and implements the open-source tool PyAnomaly for abnormal event detection,and proposes a Multi-scale abnormal event detection algorithm.First of all,based on the comprehensive investigation of the abnormal event detection algorithm in the video,this article has realized an open-source tool of abnormal event detection with high ease of use and flexibility through the improvement of the training process,the design of the project structure,and the optimization of the implementation method.This open-source tool provides researchers and developers with a unified data interface,evaluation indicators,and the realization of advanced algorithms in the industry,promoting the development of abnormal event detection.Secondly,due to the huge semantic difference between visual features and event patterns,it is very challenging to identify abnormal events in videos accurately.Existing research usually uses global features such as scenes or human appearance in the video alone to establish the connection between visual features and event types.Although this kind of method has achieved good performance,it is difficult to comprehensively judge the type of event based on the pattern of the event.Therefore,this article proposes using multiple scale information in the video to judge abnormal events jointly and offers new ideas for the use and meaning of global and local information in the event.Finally,this article explores how to use abnormal event detection algorithms to empower intelligent monitoring systems in actual scenarios.By using the method of domain-driven design,an abnormal event detection system is designed,and a prototype application of abnormal event detection in an intelligent monitoring system is designed and implemented based on Web services.