Research On Robot Path Tracking Based On Machine Vision

With the development of technology and economy,robots are gradually replacing manpower as an important force in industry.However,the low accuracy of industrial robots does not meet the highprecision requirements of end effectors.Focusing on this problem,the technology based on Machine Vision is studied in this paper,which invlove the methods of camera calibration,point matching,point registration,pose measurement and robot positioning.A prototype system tracking tools of a robot is constructed,containing a feedback between the vision and the robot.Achievements of the research can improve the absolute positioning accuracy of the robot.The research includes:I.The scheme and technical route of robot path tracking and alignment based on stereo vision,including the selection and installation of vision system,the design of software and tracking target.Firstly,the hardware of the measurement system was reasonably selected,containing cameras,lens and light sources,and the baseline and control of the cameras were determined,according to the design requirements and the results of the vision system's error analysis.Secondly,the software framework is designed.Lastly,the tracking target at the end of the robot is designed,which helps the vision system calculate the pose of robot.II.A method of camera calibration for large field of view and a method of images matching for the tracking target were proposed.The hybrid calibration method was proposed to do with the problem that the current calibration method requires a larger calibration object to get the precise parameters of binocular system with a larger field of view,which use a small-sized calibration object and a large-sized calibration rod.For image matching,the concept of Similarity was also defined,and the problem of image matching was effectively solved in the way that combines the Similarity and image segmentation based on convex hull.III.A method of global point registration based on grouping and a method of dynamic registration based on hash were respectively proposed to deal with problems of static global point registration model and dynamic point registration model,which are determined by whether there is global information of markers.The concept of Descriptor was proposed,which bases on the principle that the distance between the two points is unchanged after points are transformed by the coordinate system.The method of global registration manages descriptors by grouping,which searchs the descriptor faster,while the method of dynamic registration uses a hash map that reduces the search time,to manage descriptors.IV.A way of calculating the relative pose between the tracking target and the workpiece to be aligned was proposed,and the control of the robot was studied.Firstly,visual features were combined with mechanical by alignment structure modeling.Next,the communication between the robot and the vision system was established to get and analyze the data coming from the robot and send commands to robot.A path planning strategy was proposed for the robot motion to avoid collision.V.A prototype system of robot tracking based on stereo vision was built,and experiments verified that the end effector can align the hole in the workpiece with the guidance of the system.