Real-time people counting system using video camera

Abstract

In this MTech thesis experiments will be tried out on a people counting system in an effort to enhance the accuracy when separating counting groups of people, and nonhuman objects. This system features automatic color equalization, adaptive background subtraction, shadow detection algorithm and Kalman tracking. The aim is to develop a reliable and accurate computer vision alternative to sensor or contact based mechanisms. The problem for many computer vision based systems are making good separation between the background and foreground, and teaching the computers what parts make up a scene. We also want to find features to classify the foreground moving objects, an easy task for a human, but a complex task for a computer. Video has been captured with a birds eye view close to one of the entrances at the school about ten meters above the floor. From this video troublesome parts have been selected to test the changes done to the algorithms and program code.