Study On The Navigation Technology Of AGV With Binocular Vision System

Among the navigation technology of Automatic Guided Vehicle (AGV), vision navigation is important because of its plentiful information. The stereo vision is one of the very important parts of the computer vision, which studies how to acquire the depth information of the object in the space from two images. The thesis researches the stereo vision technique, such as: modeling of stereo vision, camera calibration technology, stereo matching technology and image processing technology. The purpose is to position the target in the space by stereo vision technology, and this technique is applied for AGV, using to examine the position of front target, and to realize positioning front target and follow. The main researched contents in the thesis are as following:1. The characteristic, the types of the AGV and the kinematics model are systemically analyzed. The method of obtaining distance information is analyzed, further more, the stereo vision model of lateral camera is researched in the thesis.2. Every method of the camera calibration is introduced. According to the actual circumstance, Tsai two marks calibration is chosen, higher accuracy and satisfied result are received by experiment.3. Stereo matching technology is analyzed, and the result of stereo matching is verified by experiment. Experiment of position the target in the space is carried out, and the result improve that the orientation of object of the vision system in the thesis is an effective and convenient method.4. The image capture card is redeveloped by Visual C++ to resemble the system of image capture and recognition. The system include program of image identifying and serial communication, which make it have the function of image capture,mark recognition and serialcommunication.5. The algorithm of target recognition based on genetic algorithm is researched. The method of similarity degree calculation by double layers model board is given. Experiment of tracking the target is carried out on the platform, and the experiment result show that the recognition algorithm is accurate,and that the vision navigation system can efficiently perform the tracking .