Research On Obstacle Avoidance Of Flat Detection Robot Based On Vision

The detection robot is a terrain mobile platform used in hazardous area to perform dangerous missions, and its independence of operation is always the research focus at home and abroad, especially in unknown environment where the ability of autonomous obstacle avoidance for the robots is needed. The motion control system of the flat detection robot is chosen as the research object in this paper and a plan of visual obstacle avoiding system is established which is based on the monocular vision to recognize and position obstacles and the path planning of obstacle avoidance based on potential field method is integrated. The feasibility of the whole plan is investigated by the means of mathematical modeling, simulation and experiment. The main work of the full text includes:The monocular vision system is selected as the medium for the detection robot to sense the environment, the obstacle positioning model of monocular vision that the camera is bias seated on pan-and-tilt is established according to the physical appearance, which effectively solves the problem that the vision field of monocular vision is limited and the imaging problem caused by the ground coordinates deviating the center of pan-and-tilt. The positioning accuracy of the model is verified by experiment.And then the image processing technology is used. The multiple obstacle recognition in color image is realized by the procedure of image threshold method, image enhancement and connectivity area abstraction. At the aspect of connectivity area abstraction, the variable neighborhood search method used to judge connectivity between the pixels is put forward, which leads the processing efficiency better than the eight neighborhood search method used before. In addition, bidirectional scanning method is adopted to scan the connection area, which helps solve the problem of markup collisions of irregular image. In order to guarantee a full scan and avoid scan transition, the number of pixels mechanism is self-designed to judge the time of repeat scan. For the location distinguish of multi-obstacle, the idea of’slicing’ managing the obstacles one by one is raised, which helps synchronize with the next-level operation and decrease wait time of the system.At the aspect of the obstacle avoidance algorithm, the artificial potential field is chosen as the local obstacle avoidance path planning arithmetic, and the control mode of underlying movement based on the artificial potential field is designed which is satisfied with the requirements of nonholonomic constraint system. Furthermore, the artificial potential field is creatively researched by simulation in the environment of ADAMS, which is better than the former research that the research object is simplified to a center point. Then the impact of the potential field gain coefficient on the obstacle avoidance path stability is analyzed selectively.At last the test platform of the visual obstacle avoiding system is set up, which verifies the practical feasibility of visual obstacle avoiding system and the relevant conclusion in simulation research.