A robust method for detecting and counting people

Estimating the number of people passing a gate or a door provides useful information for video-based surveillance and monitoring applications. This paper describes a robust method for bi-directional people counting. The method includes three steps: moving people detecting, tracking and counting. A new algorithm of detecting for moving people based on edge detection is proposed. We construct a foreground/background edge model (FBEM) from serials of frames, and retrieve the foreground edge, thus the moving peoplepsilas bounding box is obtained. Two effective methods are used for moving people tracking with the results in previous step. Besides, the counting process is described in detail, and merge/split phenomenon is also discussed to overcome the problem of people touching together. The experiment results show that a robust and high accuracy of bi-directional counting can be achieved using the method in this paper.