Research On People Counting Methods In The Video

In the video, people counting methods are the ability to access the number of openings and to provide crowd flow direction information. With the development of society, it has a wide range of application at the station, roads and the supermarket. In this paper, methods of people counting are studied in the overhead scene and the tilt scene of videos. In overhead scene, we use the background subtraction method as the prospects of detection algorithm, the design of linear statistical rule for calculating the number of the connected region that contains objects, and double regions of interesting optimize the efficiency of tracking. In tilt scene, this article use foreground detecting method based on Bayesian detection rules, and objects are extracted by the definition of a simple pedestrian model. Object tracking module uses particle filter based on Mean-shift and connected component method of combining tracking algorithm to count the number of people.