Research On The Navigation Technology Of AGV With Binocular Vision System And Lateral Track

Navigation technique is a key technique of mobile robot which decides robot's intelligent level. Vision navigation has become popular in the field of mobile robot for its great amount of information. This text takes XAUT.AGV100 as research object,in order to improve its intelligent level ,vision navigation is selected.The text expatiates double-CCD stereo vision positioning princeple, the techniques involved such as image obtaining, camera calibration, characteristic picking etc are analyzed. The aim is to make vision navigation AGV obtain its position information in the process of movement by examining the path marks. And finally it helps the AGV track the set path.The main research contents as follows:1. According to AGV development, analyzed the navigation way in common use, and did a deep research on vision navigation used in this text. The text analyzed stereo vision model and the principle of depth information obtaining,on the foundation of which the way of setting and position was determined to CCD on the AGV,meanwhile,the model of vision navigation system used in this text was built up.2. In order to build up the corresponding relationship between 2D image information and 3D space of AGV surroundings, the basic principle and characteristics of camera calibration were analyzed and compared, the two-step calibration method was selected to calibrate AGV vision system.3. Analyzed image matching technique of double-CCD vision. According to the actual surroundings and AGV's characters, in order to make it track the designed track ,two sorts of marks model were designed(characteristics matching model);brought up recognition arithmetic based on image grey level, AGV positioning for vertifying rationality of designed model and reliability of characteristics obtaining method,two different marks model were accurately recognized by stereo vision system;Positioning experiment was completed by double CCD parallax princeple for measuring distance.The experiment indicates that the designed model is rational,characteristics obtaining method is simple, viable, accurate.On the other hand,it alsovertifies the designed model can make AGV carry out fast positioning--pose estimationrelative to path marks.4. The path tracking experiments on the AGV plat-form were done. The experiment results verifies the accuracy of calibration,the selected path marks are simple and effective, characteristics obtaining method could obtain the character fast.AGV could carry out navigation by detecting path marks and the navigation accuracy satisfies the request of AGV in practice.5.The text vertifies the designed model is rational by vision system calibration, positioning experiment and path tracking experiment,the recognition arithmetic is effective and viable.The research which is done in this text is valuable for future further research.