Crowd size estimation for video surveillance

초록

Estimating crowd size in a public area is important in video surveillance applications. In this paper, a method to estimate the crowd size is proposed. The proposed method is based on a statistical comparison of the low-level image features and the size of crowd in the image. The blob sizes of moving objects and edge orientations are used as features to estimate the crowd size. To compensate for the influence of camera projection and to obtain viewpoint invariance, a feature normalization method, which considers camera projection and an elliptical human model, is proposed. Through experiments, it is shown that the proposed method can estimate the crowd size effectively. Also, experiments with different camera settings showed the robustness of the proposed method even with viewpoint changes.

키워드