Robot Arm Repeated Positioning Accuracy Measurement Using Nonlinear Local Feature Matching

摘要

A method based on KAZE feature for image registration is proposed in robot arm repeat positioning accuracy measurement method. Repeatability testing on the XY-Theta test platform. The KAZE feature extraction and matching are performed on the three sets of images taken by the industrial camera, and the offset coordinate point cloud is obtained, and the minimum enclosing circle of the point cloud is fitted, thereby obtaining the repeated positioning accuracy of the mechanical arm. The comparison test results show that the proposed method achieves image matching accuracy of less than 0.01 pixels, which is superior to SIFT and SURF based methods in accuracy and stability. Furthermore, the repeated positioning accuracy (ISO experimental average) of the XY-Theta experimental platform under coarse positioning and fine positioning conditions was 2.38052μm and 0.67628μm, and the scheme achieved satisfactory results on the XY-Theta platform.