Crowd Behavior Analysis and Classification using Graph Theoretic Approach
Surveillance systems are commonly used for security and monitoring. The need to automate these systems is well understood. To address this issue we introduce the Graph theoretic approach based Crowd Behavior Analysis and Classification System (GCBACS). The crowd behavior is observed based on the motion trajectories of the personnel in the crowd. Optical flow methods are used to obtain the streak lines and path lines of the crowd personnel trajectories. The streak flow is constructed based on the path and streak lines. The personnel and their respective potential vectors obtained from the streak flows are used to represent each frame as a graph. The frames of the surveillance videos are analyzed using graph theoretic approaches. The cumulative variation in all the frames is computed and a threshold based mechanism is used for classification and activity recognition. The experimental results discussed in the paper prove the efficiency and robustness of the proposed GCBACS for crowd behavior analysis and classification.
