Moving Obstacle Avoidance of a Mobile Robot Using a Single Camera

Abstract

This paper presents some preliminary results of the detection of moving obstacles (particularly walking humans) by the use of a single camera attached to a mobile robot. When a camera moves, simple moving object detection techniques for a stationary camera, such as background subtraction or image differencing, cannot be employed. We thus detect objects that move near the robot by block-based motion estimation. In the method, an image is firstly divided into small blocks, and then the motion of each block is searched by comparing two consecutive images. If the motion between matching blocks is significantly large, the block in the current image is classified as belonging to moving objects. The method is verified by the indoor navigation experiments of a robot.