Motion Pattern Recognition from Crowded Video

During the previous decades, in the era of video surveillance, an immense consideration had been given to crowd video control and crowd management. Among the undertakings for video surveillance, retrieval of crowd video is a very challenging task. Motion Pattern Recognition from Crowded Video generally faces the problem of occlusion, low resolution, cluttering sometimes object social interaction is not present then find the abnormal behavior is quite tough. We handle these pattern recognition problems by introducing the social force model approach in which we calculate the social force between pedestrians presented in the crowd which helps in recognizing the behavior of the crowd in an abnormal situation. In our approach, we find some well-known types of motion pattern these are: lane, ring, fountainhead, and bottleneck [1]. In this paper, our proposed approach focused on some steps these are: preprocessing of frames, calculate optical flow, calculate the social force model among pedestrians, Clustering after Jacobian matrix calculation and in the last step behavior recognition with the help of KNN classifier.