Research On Parts Recognition And Orientation Algorithm For Convey Robot Based On Stereo Vision

With the advent of industrie4.0 age,the scale of the robot industry is rapidly expanding.Gradually,traditional human production mode will be replaced by robots'.Especially,robot has been widely used in the hard and repetitive transit works.However,the convey robot complete the request work with the method of artificial teaching,off-line programming or two-dimensional vision guide.In those way,the position and orientation of target was needed to ensure in advance.Combining the stereo vision and robot handing system can tackle these problems effectively and improve robotic autonomy and flexibility.In the paper, the key techniques of stereo vision handing system,such as Image Matching,target recognition and three-dimensional pose estimate,are researched and demonstrated, in the paper.Four aspects of works have been completed as following:Firstly,three-dimensional visual handing system was constructed with single binocular vision method based on the configuration of “eye in hand ”.The camera calibration algorithm and hand-eye calibration method was researched on the basic of introducing the camera imaging model.and the calibration experiments was carried out and the results were analyzed.Secondly,designed the recognition method of weak texture target by combining improved shape descriptor,contour moment feature and sub contour feature.and it is proved that the method can accurately identify the target of the weak texture objects.And then,the Fundamental matrix estimation algorithm and the 3D pose estimation method was studied. First,the traditional NCC function was optimized by the difference principle and the region expansion matching strategy for improving the speed of points matching.Second,fast clustering algorithm was used to purify the matching points for improving the accuracy of matching.With those matched data,the Fundamental Matrix was estimated accurately by nonlinear iterative algorithm which improved by statistical theory.Third,the key features of objects were matched under the constraints of epipolar geometry and the 3D position of target was confirmed based on Fundamental Matrix.Finally,the experimental system of parts recognition and orientation which consisted of IRB-1410 ABB robot, CCD camera, end effector, PC, PLC and other components was established.The experimental results show that the algorithm is effective.