High-accuracy 6-D pose measurement method for 3C thin parts in robotic assembly by monocular vision

The 6-D pose measurement of the 3C thin parts is essential for alignment process in robotic assembly. Due to the thin thickness, features from two surfaces are coupled together and difficult to distinguish in the image, causing the decrease in the accuracy of the pose measurement. In this study, a 6-D pose measurement method based on the monocular vision is proposed. An algorithm named dual-surface reprojection contour error optimization (DSRCEO) is proposed to simultaneously optimize the features from both two surfaces to improve the accuracy of pose measurement. In the DSRCEO, the computational domain and the image domain are constructed, and then an error index of DSRCE is derived by comprehensive consideration of the two domains to evaluate the quality of the current pose estimation. By minimizing the DSRCE, the initial estimation of the 6-D pose is iteratively optimized to continuously improve the measurement accuracy. The framework of the DSRCEO is available for the thin parts with various shapes, and specific algorithm implementations for the two most common shapes (trimmed circle and polygon) are derived. Finally, the accuracy and applicability of the proposed algorithm are verified through both sufficient simulation and experiments.