Research On Robot With The Function Of Binocular Ranging

With the continuous development and breakthrough of computer hardware and software, binocular ranging based on image processing has become a hot topic. Binocular ranging has changed the traditional measurement methods, which can be used in many fields of unmanned, 3D reconstruction, intelligent control. The amount of information collected by the vision sensor is large and wide, it is called the most important sensor of the robot, if visual sensor is applied to a robot, we will get more interactive information to carry out task effective, and the traditional model of robots will be changed. Therefore, the research on robot with the function of binocular ranging is studied, which not only has important theoretical significance, but also has good market prospects and wide application value.In this thesis, based on the Linux development environment, a robot with the function of binocular ranging was studied and designed after the analysis and research of video frame capture, image preprocessing, camera calibration, image matching, re-projection and other theoretical methods. The robot collects the video information about the surroundings, and the information are processed by Mini6410 to obtain the 3D point cloud of the visual angle, then after the analysis of 3D point cloud the depth of the closest object is concluded. At the same time, the video information collected by robot are coded by the server, and then the compressed information are transmitted to the android client, and the staff can control robot for specified tasks based on the video image.This thesis mainly completes the following several aspects:(1) Study on the control system which has the hardware platform of MCU. Firstly, the control principle of traveling mechanism and the control principle of manipulator are analyzed. Then, based on the analysis of serial communication, interrupt and other knowledge, a reasonable design of the hardware structure of the robot is designed, and the control design of the robot is completed.(2) Based on pinhole imaging model, the camera imaging principle are analyzed, the points’ transformation between the image pixel coordinate system and the world coordinate system is established, and the intrinsic parameters of the camera are obtained by calibrating the camera with a known size of board, finally, stereo calibration is achieved by using the principle of epipolar constraint, Marr constraint and so on.(3) Based on the analysis of the basic algorithm and principle of image matching, the matching feature points are found by using SAD algorithm to match the left and right images, and the 3D point cloud are obtained by using the feature points’ disparity between the left and right image, then, the depth of the closest object is concluded after the analysis of 3D point cloud.