Video Based Crowd Density analysis for Visual Surveillance

Abstract: Crowd density estimation is the challenging task for visual surveillance. Due to presence of high risk of degeneration, the safety of public events has large crowds and it has always been major concern. Video analysis techniques are becoming increasingly popular in the visual surveillance of public areas because of their great efficiency in gathering information and low cost in human resource. Its central topic is automatic analysis and detection of abnormal events. In monocular image sequences, the combined method of foreground based, feature based and group based methods is applied to extract crowd areas having motion. The specific number of persons and velocity of a crowd can be adequately estimated by the system from crowded areas. Having a camera network, crowd density can be predicted with the system. The system has lot of been applications such as in real applications, and experiments conducted in real scenes (station, college) demonstrate the effectiveness of this method.

Keywords: crowd density, human tracking, Haar based features.