Abstract: The real-time calculation of positioning error, error correction and state analysis is a difficult challenge in the process of end-effector autonomous positioning. In order to solve this problem, the Kalman filtering based on three-frame subtraction is proposed to capture the moving end-effector. Back propagation (BP) neural network is adopted to recognize the target. And 3D information extraction based point cloud library (PCL) is described to calculate the space coordinates of the end-effector and the target. The scattered points are processed by gridding and interpolation. Experiments demonstrate that the end-effector positioning can be corrected in a short time. The prediction accuracy of position reaches 99% and the recognition rate of 99% is achieved for target object. Furthermore, the gradual convergence of end-effector center (EEC) to the target center (TC) shows that the autonomous positioning is successful. The algorithm effectiveness is also validated by 3D fitting, but the camera calibration is not required. The system efficiency is improved.

Key words: autonomous positioning, surface fitting, end-effector, back propagation neural network, point cloud library