A Heuristic Information Based System for People Counting

摘要

In this paper, we present a people counting system by accumulating the moving pixels in a region of interest (ROI) called Virtual Gate frame by frame. The proposed heuristic method is based on pedestrian detection and is used to determine the proportion of each point on-gate to the pedestrian model at its position. The amount of people can be abtained by adding up all those on-gate weighted moving pixels. We validated our system on a public dataset and an 8 minutes long video taken by us, of which the cameras' tilting angles are in special, and promising counting results were got.