Resarch About Obstacle Detection And Avoidance Of Mobile Robots Based On Monocular Vision Technology

Computer vision technology is at a stage of rapid development nowadays,which can be employed to realize distance measuring, objects matching and geometric modeling by using processing algorithms on images captured. When this technology is applied to obstacle avoidance research upon mobile robots,compared to binocular vision technology, monocular vision technology has a faster algorithmic speed when dealing with vision processing task during obstacle avoiding of mobile robots, as well as an improved real-time performance. This thesis is about research about obstacle detection and avoidance of mobile robot based on monocular vision technology.The obstacles in the video which is captured need to be detected by single camera. The obstacle detecting algorithm in this thesis is based on Frame Differential Method and combined with Pyramid LK Optical Flow Method. Compared with Frame Differential Method, this algorithm can detect fast moving obstacles; compared with Pyramid LK Optical Flow Method, this algorithm has smaller computing region and less computations, which can offeran faster algorithmic speed and can better satisfy the real-time requirement of obstacle avoidance system of mobile robot.Robots need to make decision after completing obstacle detecting to avoid knocking on the obstacles. This thesis uses VFH obstacle avoidance algorithm by analyzing the motion model of mobile robot. This algorithm makes decision of robot’s moving direction of next moment by measuring probability of existence of obstacles in the grids and distance between robot and obstacle at present moment: robot turns left or right according to the position of obstacle when there is an obstacle in front of the robot and the distance between robot and obstacle is less than threshold; robot goes straight when there is an obstacle in front of the robot and the distance between robot and obstacle is larger than threshold; robot goes straight when there is no obstacle in front of the robot.To prove the validity of the algorithm and method adopted in this thesis, a complete verification platform is established. The ultrasonic sensor on the NI start package was replaced by a web camera AXIS 206 in front of the NI start package to realize the data transmission between LabVIEW and web camera;And the router ASUS RT-G32 is used to realize wireless communication between computer, web camera and NI start package; Power supply of web camera and router was completed on NI platform. In addition, the software LabVIEW was employed to build obstacle avoidance system, which includes vision acquisition algorithm, obstacle detection algorithm and decision making algorithm, and experiment was conducted on the visual platform. The final resultshows that the obstacle avoidance algorithm in this dissertation is accurate and fast, which can satisfy the real-time requirement of system and is of important practical value.