Pedestrian detection by fusing multiple techniques of a single color camera for a mobile robot

Abstract

In recent years, many electronic/mechanical systems have acquired mobility due to rapid technological advances and market demands. Mobile robots are a typical example. When a robot moves around in an uncontrolled environment such as a building corridor, safety of the robot and humans around is of great importance. Although ultrasonic range finders have been popularly employed so as to sense obstacles in front of a mobile robot, they have many limitations due mainly to low sensing resolution in such a narrow space. In this paper, we present an effective method of pedestrian detection by using a single color camera mounted on a mobile robot that navigates in a corridor. Vision techniques of detecting appearance and body are practiced and their results are fused to obtain more reliable results.