Research On Technology Of Mobile Robot Obstacle Detection Based On Binocular Stereo Vision

Mobile robot obstacle avoidance technology is active branch in the field of robot Research, this technology is also Multidisciplinary comprehensive subject. Obstacles detection is a prerequisite for obstacle avoidance safety. Vision sensor has incomparable advantages compared to other sensor. For this reason, Mobile robot obstacle detection technology based on vision has great research value and practical significance.In this paper, binocular stereo vision technology was used for obstacle detection on the path of indoor mobile robot. The detection step for: image acquisition, camera calibration, stereo rectification and matching, obstacles extraction and ranging on the path of Mobile Robot. According to these four parts, this paper conducted an in-depth theoretical analysis and experimental research, and builded a minitype experimental platform of binocular stereo vision. The main research were as follows:1. Camera Calibration: Firstly, Camera imaging model was introduced, then linear calibration method, Tsai two-step calibration method and Z Zhang plane template two-step method were studied in detail. Under the camera distortion model, the simple and precise Z Zhang plane template two-step method was selected to calibrate the binocular camera. Through the projection error analysis, the calibration results were verified correct.2. Stereo Rectification and Stereo Matching: Bouguet rectification algorithm was studied and used in actual scene images based on Open CV vision library. In stereo matching, block matching algorithm and semi-global matching algorithm were studied. At last, semi-global matching algorithm was selected for actual scene images because of its high speed and precision.3. Obstacles Detection on Path of Mobile Robot: Obstacle detection method based on disparity map was studied, it was divided into four steps: the first step was the three-dimensional measurement based on the disparity map, the second step was removing the region outside the path of the mobile robot. the third step was removing the surface interference, the last step was performing binary segmentation within a given distance range. Using this method, obstacles on the path of mobile robot was detected successfully.