Development and Calibration of a Cross-Scale Multicamera Vision System for 3-D Microassembly

Abstract

Focusing on the requirements for large-scale positioning and high-precision operation in various microassembly processes, this article proposes a cross-scale multicamera vision system and a corresponding calibration method. In this system, binocular global vision (BGV) and binocular micro vision (BMV) with clear imaging volumes (CIVs) of different scales are integrated, where the CIV of BMV is nested in that of BGV. To achieve the joint calibration of the BGV and BMV, a cross-scale calibration board (CCB) is designed, featuring a global ChArUco pattern embedded with micro chessboards. In addition, a three-step calibration strategy is proposed. First, the BGV and BMV are calibrated independently using the global and micro features on the CCB. Second, the global and micro corner points are reconstructed in 3-D, respectively, and the prior geometric constraints of the CCB are used to achieve the coarse calibration of the cross-scale visual system. Finally, the global information is integrated to optimize the BGV reconstruction results within the limited common CIV of all cameras, thereby achieving calibration parameter optimization in the world coordinate space. A series of experiments were conducted to verify the effectiveness of the designed CCB and the proposed calibration algorithm. The results show that compared with the traditional chessboard, CCB has obvious advantages in speed, accuracy, and ease of operation, and the calibration algorithm can effectively reduce the alignment error between global and micro CIV in cross-scale vision systems. Finally, the effectiveness of the calibration algorithm was verified through two application scenarios involving collaborative assembly robots.