Abstract
Locating all obstacles around a moving robot and classifying them as static obstacles or not by a sensor are essential for the robot's smooth movement and avoid problems in calibrating many sensors. However, there are few works on locating and classifying all obstacles around a robot while it is moving by only one omnidirectional camera. In order to locate obstacles, we regard floor boundary points where robots can measure the distance and direction from the robot by one omnidirectional camera as obstacles. Tracking them, we can classify obstacles by comparing the movement of each tracked point with odometry data. Moreover, our method changes a threshold to detect the points based on the result of comparing in order to enhance classification. The classification ratio of our method is 85.0%, which is four times higher than that of a method without changing a parameter to detect the points and that of a previous method.
