Mobile Robot Obstacle Avoidance Binocular Stereo Vision Algorithms

Tracked Land Robot is an intelligent mobile robot, which can run autonomously, and continuously on road or cross-country. Its research which embodies the latest achievements of information and artificial intelligence, is of great value in research and application. Of all the Tracked Land Robot key technologies, one of vision-based navigation's functions is the rapid detection and recognition of obstacles. Compared to other approaches, stereo vision is rapid, accurate and easy to conceal, as it is a passive measurement, which does not emit light and radiation. Therefore, stereo vision navigation which based on avoiding obstacle is a very important part of the Tracked Land Robot research. Robot vision, especially stereo vision, is involved in several different projects in my lab. To fulfill the requirement of these projects, we focus our research on stereo vision techniques, especially stereo matching and obstacle detection etc.This paper primarily studies the stereo vision matching algorithm which based on avoiding obstacle. The process is as follows: image calibration, image smoothing preprocessing, image matching, background and obstacle separation, obstacle 3D reconstruction, extracting obstacle. This paper proposes a non-parametric transforms stereo matching algorithm which in view of image matching. We use multi-resolution stereo matching, and multi-windows matching in lower resolution image. The result will support matching windows size and the disparity range for the higher resolution image, and distinguishing background and obstacle will improve the performance of matching. In extracting obstacle phase, the proposed method is use chain-code tracing to trace the possibility area of the obstacle, then wipe off the false area and get the true area of the obstacle. Obstacle 3D reconstruction gets the position and distance of the obstacle.The whole algorithm of stereo matching which based on avoiding obstacle described in this paper is flexible, real time and exact, and not designed for any special robot system. As a consequence, this algorithm can be applied to any other stereo vision system. Our stereo vision system shows very good performance on obstacle detection in experiments. Therefore, this thesis can be the foundation for future related work.