An Adaptive Swarm Optimization Technique for Anomaly Detection in Crowded Scene

Since last decade, crowd behaviour analysis and management gained lots of consideration from the researchers for the intelligent video systems. Automated surveillance systems faces challenges in crowd behaviour modeling and analysis because of dynamic characteristics of crowd and individuals. In this paper we propose a new approach for the detection and analysis of crowd behaviour by using adaptive swarm intelligence and optical flow estimation based approach. According to this approach, initially image is modelled to generate the optical flow. This modeled image contains foreground, background and image region (higher intensity). Optical flows and streaklines are used to represent motions observed. The observed motions are analyzed using particle swarm optimization. The simulation study is carried out on the publicly available dataset from University of Minnesota using MATLAB simulation tool. Experimental study shows that the proposed approach is more efficient when compared to the existing approach for the detection and behaviour analysis. Comparative study is carried out in terms of classification error and area under curve.