A Method for Counting People at Elevator Lobby Focusing on Top of Moving Objects

Abstract

This paper proposes an image processing system for counting people at an elevator lobby viewed from an upper oblique angle. The method is based on two assumptions: heads of people exist around the upper edge of the images of moving objects; and shape of the head is circular. This paper shows that the method based on these assumptions acquires high recognition rate and high reliability in an elevator lobby with many overlapped images of people.