Visual tracking of human back and shoulder for person following robot

Abstract

This paper presents a tracking vision for human collaborative robot. It tracks human back and shoulder for the robot to follow a person. Obviously it would be very convenient and comfortable for us if a robot could follow us in the real world. But it is not easy to keep tracking human back even if the background image is so complex and cluttered. Thus the authors solve the problem by utilizing their robust pattern detection and identification system. It is robust against fluctuation of pattern location, size, and brightness. It contributes to robustness of the tracking that the authors choose texture of clothes and human shoulder image as template patterns to be detected and identified. Experimental results demonstrate feasibility and effectiveness of the idea.