The Stereo Vision Technique Used For The Mobile Robot

The problem of motion planning is very important for the mobile robot. In order to solve the problem, the mobile robot must understand the environments. The robot's vision is very important for the robot to understand the environment. The stereo vision is one of the very important part of the computer vision, which studys how to acquire the depth information of the object in the world from two images. This thesis is supported by Tianjin's Natural Science Foundation "The Study of Mobile Robot System Technology Faced to Complicated Task"(No. 003601211) and Hebei Province's Tackling Key Problem Plan "The Study of AGV Accurate Navigation Used in Material Automation Transportation"(No. 00547001D-18). This thesis mainly focuses on the stereo vision technique used for the mobile robot, the main work includes:1. The HEBUT-I mobile robot's structure and main hardware are introduced, at the same time, the features of the speed and displacement of the HEBUT-I mobile robot's driving system are analyzed quantitatively. On the base of quantitative analysis, the driving mode is defined. The vision system of HEBUT-I mobile robot is designed and the function of every parts is introduced.2. Based on the analysis of the Pin-hole model, the model of orthogonal projection and the model of approximate projection, for HEBUT-I mobile robot, the Pin-hole model is more adapt one. According to the installation of the camera on the HEBUT-I mobile robot, analyzing the stereo vision model of lateral camera in detail in this thesis. At the same time, common visible region and the model error of this model are studied deeply, further more, the error produced by the error of the correspondence is analyzed in detail.3. Based on the analyzing of every method of the camera calibration, according to the requirement of the orientation of object, anew method for camera calibration quickly is put out. The results of experiment shows that this correction method is robust and accurate, which can satisfy the need of the orientation of the vision system of mobile robot.4. Because the actual work environment of the mobile robot is complexity, on the base of analysis of several edge extract operators, the Sobel operator has been selected to accomplish the edge extract in this thesis. A two-stages orientation of object has been developed, which is general orientation firstly, then from this orientation point expanding to search the edge of theobjective region. It can ge the center of the image after extracting the image edge. This method can greatly reduce the computer quantity of the feature extract and stereo matching and then improve the speed of recognition.5. The results improve that the orientation of object of the vision system in this thesis is an effective and convenient method.