Obtaining Obstacle Information by an Omnidirectional Stereo Vision System

This study describes a method to obtain obstacle information omnidirectionally by a novel omnidirectional stereo vision system. Such information is important for mobile robots navigation. The omnidirectional stereo vision system is based on a common perspective camera coupled with two hyperbolic mirrors. Given two images of a space point acquired by this vision system, the 3D coordinates of the point can be calculated by means of triangulation. Then the study presents a full model of calibrating this vision system, which can be easily generalized to all kinds of catadioptic sensors. The images captured by the stereo vision system are unwrapped into panoramic images, which are then examined for stereo matching along vertical epipolar lines. An algorithm of the quick stereo edge matching method based on Canny edge detection is proposed. In the end we show the result of obstacle detection.