Based On Binocular Vision Technology Of Mobile Robot Navigation

With the rapid development of the machine vision technology and robot navigation technology, the robots with a higher intelligence and stronger adaptive ability are the development trend in the practical application of robots. Therefore, the mobile robot navigation technology based on the vision system is one of the hottest focus issues in the field of machine vision. Location and target recognition technology are the foundation and key to achieve successfully the precise navigation for robots. A comprehensive scene description and a wealth of image information are the important sources for mobile robot that perceives the environment. In recent years, vision-based navigation technology has been widely used, but the problem of real-time still needs solving in the practical application. Combined with the practical application, the real-time images information is obtained by using binocular vision system that simulates the structure of the human eye. The camera calibration, image registration, target recognition and hardware design based on embedded system all are further analyzed and researched for mobile robot navigation in the thesis.The research purposes and values of the mobile robot navigation based on the binocular vision, as well as the composition of binocular vision system are first discussed. Then the developed states at home and abroad and key technologies are introduced for the mobile robot navigation technology. According to theory of binocular stereo vision, the image is obtained by using the binocular cameras which films the same scene in the different positions. Camera calibration is the foundation of binocular vision system. Compared with the camera model and the typical calibration method, Tsai two-stage that is used to achieve the camera calibration can effectively obtain the internal and external parameters of the camera and complete the high precision calibration for the binocular camera. The method can achieve the precise location of the object. Image registration is important steps for the image processing and analysis, which is the necessary condition for target recognition. The corner features are extracted by using fast corner detection method. The rough matching corresponded points are achieved by computing the correlation of bi-directional grays. Then the precise matching is achieved by using the epipolar constraint and the fundamental matrix. Experiments show that the accuracy and speed of stereo image matching can be improved by two-steps matching. In addition, target recognition based on the shape achieves the accurate position for mobile robot. Finally, mobile robot is moved by path planning. Therefore, the binocular function of mobile robot navigation is achieved. The fast processing of larger image information and the data of rapid computation are achieved by using ARM embedded with the core processing chips and FPGA as an assistant coprocessor in the hardware design. The efficiency processed and the speed run are improved by the Optimization to solve the real-time problem in the mobile robot navigation technology.