Research On Robot Dynamic Obstacle Detection And Obstacle Avoidance Method Based On Monocular Vision

As one of the most important research techniques of a new generation of technology revolution in robot technology. As in the field of the industrial robot robotics research has earlier skilled by people, it has been widely applied to many fields of production, processing, manufacturing etc. In the industrial robot operating environment, how to avoid the robot path at any time into the staff or other dynamic obstacle collision, has become one of the important topics in the field of robot safety operation.Starting from the robot operation angle, in the robot installation environment information acquisition device, analysis and monitoring of the surrounding environment, so as to avoid the occurrence of such accidents is very necessary.According to the above mentioned problems man-machine safety operation, using the small mechanical arm with ESPON six degrees of freedom at the end of the installation of the existing single camera laboratory as the research object, and build a robot in the specified path to the target object moving along, random experimental scene dynamic obstacles. The model estimation of motion parameters on the dynamic environment to build the global use of the robot’s visual information, combined with the background of compensation technology, realize the separation of background and foreground in dynamic scenes, due to the elimination of the end effector of a robot motion feature extraction of moving obstacle influence; Secondly, through the analysis of the optical flow algorithm, using an improved calculation method of KLT feature point tracking based on optical flow, realized the robot to detect moving obstacles in the path of motion; Finally, the position of the obstacle at the next moment is obtained by optical flow computation, according to the collision time of obstacles to calculate the depth information of obstacles, combined with the artificial potential field method, the virtual repulsive force and gravity force between the robot and the object and the object are constructed, so that the robot can avoid obstacles and close to the target. In this paper, a method based on optical flow is used to detect the obstacle avoidance of the robot with artificial potential field theory, with a single camera as the image acquisition equipment, can not change the existing automated production line visual servo robot on the premise of any hardware structure, the realization of the robot’s autonomous obstacle avoidance. At the same time, based on binocular vision for obstacle detection compared to the system by the speed of image optical flow estimation of the relative distance between the robot and dynamic obstacles, effectively avoid due to the site environment, light effects or texture lack in binocular image matching fails to better ensure the robot to avoid the avoidance of reliability and real-time.This paper mainly completed the following works:1, Using the robot end effector camera scene dynamic image information extraction, and to realize the separation in dynamic scenes of background and prospects through the dynamic image information parameters of global motion estimation and background compensation; motion parameters of the camera model is established, on adjacent inter frame motion compensation is performed by solving the parameters of the model, eliminating the robot end effector motion of obstacle feature extraction effect;2, Using the improved calculation method of KLT feature point tracking based on optical flow, realized the robot to detect moving obstacles in the path of motion.3, Through the calculation of the obstacle feature point optical flow, get the obstacle of position estimation, extraction of obstacle of relative depth(Time-To-Contact) information, and the artificial potential field theory and the obstacle depth information to combine, robot dynamic obstacles to avoid and close to the target object.4, Using Visual Studio 2010 programming software, the improved implementation of KLT feature point tracking optical flow computation and obstacle depth information extraction based on, on this basis, the use of visual information of the robot dynamic obstacle detection and avoidance test fully validated the avoidance obstacle detection and obstacle avoidance method is effective and feasible.