DETECTION OF UNCONTROLLED MOTION BEHAVIOR IN HUMAN CROWDS

Analysis of crowd behavior using surveillance videos is an issue for public security. Crowd behavior can be analyzed using two approaches, first one is analyzing individuals behavior in a crowd which is termed as object based and the other one is analyzing the crowd as a whole which is termed as holistic based approach. Identifying and tracking the crowd motion is a major interesting research topic in vision analysis, crowd dynamics and visual surveillance fields. In order to develop the proposed work holistic approach is used. Crowd is the agglomeration of many people in the same area at same time. Music festivals, sports events, pilgrimage are the mass events. In such mass gatherings, the density of the people is more and this may often lead to crowd disasters. Due to this, people will die by suffocating high pressure on their chest. This system is developed to overcome such situations. The proposed system will help the security personnel to evaluate the situation to take necessary actions against the crowd disasters. The decision “if and how to ” react solely based on the expert knowledge of the security personnel. The proposed system is based on the optical flow computations and detects the patterns in crowd motion that indicate the dangerous situations. Lucas – Kanade method is used to find the optical flow. Here two input videos are used to conduct the experiment. System will allow only authorized person to use it by entering the “Username and Password”. The videos consist of many numbers of frames and this system will extract those frames in the input video. After extracting the frames from the input video, it will divide each frame into sub parts for easy and efficient processing of video sequences and for the respective subpart optical flow vectors ar