Abstract: When conducting segment assembly, the position and orientation of the segments should be adjusted to align with adjacent segments after transporting to the designated position. To accurately capture installation position and orientation of the segments,, the authors develop a method for rapid identification of the posture of installed segments and prediction of the posture of segments to be assembled. The method employs a structured-light camera to capture the longitudinal joint point cloud of installed segments and applies a two-stage hand-eye calibration to transform the point cloud into the assembly robot′s base coordinate system. Based on the segment point cloud, longitudinal joint segmentation and circumferential joint detection are performed to determine the posture of installed segments. Using the arrangement scheme of the segments to be assembled, their preliminary installation positions and orientations are derived. An on-site experimental platform for automated segment assembly is constructed, and assembly tests are conducted to validate the proposed algorithm. The results show that: (1) the developed longitudinal joint segmentation method, based on density clustering and a support vector machine, effectively extracts the midline of the longitudinal joint; and (2) the proposed positioning method achieves low errors, with coordinate deviations in all three directions within 12.3 mm, thereby meeting the requirements for coarse positioning in segment assembly. This method provides a reliable initial reference for subsequent fine adjustments of segment posture.

Key words: automated segment assembly, coarse positioning, machine vision, structured light, posture detection, assembly machine