Human positioning based on probabilistic occupancy map

With the development of video analytics technology, Video surveillance system can detect and report suspicious objects location information in time, which gradually changes from post-mortem forensics to the pre-prevention mode. By defining a discrete occupancy map to describe whether the object is in the certain space, the algorithm in this paper iteratively estimates the occupancy probability for each location so as to quickly determine the specific position of the moving objects in the video. Videos from Internet and taken in the lab are both tested in this paper. The result shows that this method can effectively solve the problem of occlusion and information fusion of multiple camera. When monitoring the scene of 3 × 3 (m), the positioning accuracy is over 96%.