Pose measurement approach based on two-stage binocular vision for docking large components

Abstract

Large components docking is an important step in industrial manufacturing. During the docking process, the relative pose of joining components needs real-time and accurate acquisition. Existing methods rely on expensive and complex optical instruments. Photogrammetry has the advantages of low cost and fast measurement speed, but its measurement accuracy decreases sharply along with the increase of the measurement range. This paper proposes a two-stage binocular vision which consists of two sets of binocular vision systems with different accuracy levels in the same coordinate system. Binocular system with corresponding structural parameters is designed for pose measurement at corresponding stages to solve the contradiction between the range and accuracy in measurement. A triangulation and spatial plane fitting method is proposed to calculate relative poses without introducing coordinate transformation errors. We adopt a novel 3D optimization method to further improve the accuracy. Experiment results show that this method can meet the measurement range and accuracy requirements for large components docking. Compared with traditional combined vision measurement station based on multi-vision sensor, the proposed method reduces coordinate transformation errors and overcomes the contradiction between the measurement range and accuracy, which can improve the accuracy and meet the requirements of engineering application.