Human Crowded Scene Analysis for Behaviour Recognition Using Nature Inspired Algorithms

Abstract

Activity recognition gained immense popularity due to the increasing number of surveillance newlinecameras. In addition, identifying the actions from crowded video sequences is the prime newlinefunctionality of intelligent video systems. The purpose of activity recognition is to detect the newlineactions from the series of estimations by varying the environmental condition. However, the newlinemethods based on conventional features utilize pixel information which faced several issues, newlinelike dynamic backgrounds and illumination changes. However, panic behaviour is also the newlinenext level of abnormal behaviour in a human crowd, so detecting panic behaviour helps newlineprevent disastrous situations. Various existing methods are adopted for detecting panic newlinebehaviour in crowded scenes, but it results in performance degradation due to the varying newlinedensity of the crowd. Violent crowd behaviour detection has increased major attention in newlineComputer Vision systems. Diverse crowd behaviour detection approaches are introduced to newlinedetect violent behaviour but enhancing the recognition rate poses a complex task due to newlinedifferent crowd diversity, mutual occlusion between crowds, and diversity of monitoring newlinescene. newline