Object Detection And Positioning Of Mobile Robot Based On Binocular Vision

Human get outside information is obtained 80 percent by visual, with the extension of the application of robots and the development of computer technology, it is hoped that the system of robot can have similar to the human eyes to observe the surrounding environment, replacing human in dangerous situations such as severe pollution area to work. Binocular stereo vision system can be used to simulate the eyes of people to deal with the objective scene, it is about two parallel arrangement of cameras to obtain the image of the outside world respectively, to deal with two images to obtain 3D information from the outside. Positioning is considered as the premise and foundation of mobile robot autonomous navigation, path planning and obstacle avoidance tasks. Mobile robot only know where its location, it can purposefully exercise and perform specific tasks. Nowadays mobile robot positioning based on binocular vision usually needs help by detecting and extracting artificial landmarks, however, because of the limitation of mobile robot’s movement and binocular vision’s field, tends to appear the landmark unable to detect or do inaccurate detection. To solve this problem, a method, by which the self-positioning of stationary work robot was firstly performed and then the moving object was detected and positioned, was proposed. The main content is divided into the following several aspects:First of all, the binocular stereo vision system of mobile robot was studied. After introducing the mobile robot voyager II and binocular cameras Bumblebee2, we studied the key technology of binocular stereo vision system and the principle of binocular stereo vision parallax, and got the transformational relationship between binocular stereo vision coordinate systems by the camera model, and transformed from coordinate values in two-dimensional images to three-dimensional coordinates in the space accordingly, to be paving the way for the positioning calculation of mobile robot.Secondly, aiming at the static object in the environment, detection step, including selecting an appropriate color space, morphology processing, connected component tag and image segmentation, etc., was proposed, and the artificial landmark in the experimental environment was detected and extracted. For detecting moving object, interframe difference method was used in this paper, which can extract the outline of mobile robot in the experimental environment completely.Finally, during the course of the experiment the moving object performed the self-positioning based on artificial landmark, the circumstance of precision of self-positioning of the moving object not high was found. In order to improve the precision of moving object positioning, this paper a method, by which the moving object in the process of moving needs stationary work robot to help it perform the positioning, was proposed. The stationary work robot was firstly performed the self-positioning and then the moving object was detected by frame differential method and positioned by projection location method. The result of experiment proved that the method of moving object with the aid of work robot performed the positioning had higher precision than the method of traditional moving object performed self-positioning, and the algorithms have good real-time performance, so this positioning method is suitable used in the situation of high positioning precision and it has practice significance.