Abnormalities detection and tracking of crowd using kinematics approach in a video stream

The effective analysis and detection of crowd behavior is a very challenging issue in video surveillance systems. Most of the conventional approaches is not up to the mark for real time behavior analysis. In this work, an optical flow method along with KNCN classification approach is proposed to recognize and track abnormality in a video scene. Optical flow is an efficient global motion information algorithm. The optical flow vector provides direction and displacement of a point between two consecutive frames. This feature vector acts as input for the classification stage. A surrounding neighbor (SN) concept is used for finding k-nearest centroid neighbors (k-NCN) for each frame, where the class level is assigned based on the votes of k-NCN. The simulation results are promising and seems very efficient in terms of accuracy. The simulation results are analyzed and a frame level comparison with the ground truth is performed. ROC analysis provides area under curve (AUC) for different data sequences and is compared with other existing techniques; moreover, our proposed model KNCN-KADA obtained a promising accuracy of 99%.