High Dense Crowd Pattern and Anomaly Detection Using Statistical Model

Human crowd behavior analysis is a subject of great interest in research now days. Great advantage of investigating dense human crowds in places like mosques and temples to perform automatic surveillance for any unusual activity detection that might be a subject of interest and must be addressed on earliest to avoid accident. We present robust statistical skeleton for modeling a dense crowded scene and then find anomaly in it. Our main intuition is to utilize the true dense motion of the scene to model main motion pattern by using temporal statistical behaviors. After removing the noise we genuinely find anomalies in the scene by using the model. We used Matlab for designing algorithm of motion pattern and anomaly detection. Our test reflect that temporal motion pattern modeling presents hopeful results in actual world scene with dense structured crowded motion.