摘要 : Due to the advantages of high accuracy and rich informa-tion, point cloud-based geometric reconstruction has been widely applied in 3-D reconstruction for various scenarios. However, the classic iterative closest point (ICP) algor... 展开 Due to the advantages of high accuracy and rich informa-tion, point cloud-based geometric reconstruction has been widely applied in 3-D reconstruction for various scenarios. However, the classic iterative closest point (ICP) algorithm, relying on threshold permission, imposes limitations on the distance and generalizability of point cloud registration. In this letter, we propose a 3-D reconstruction system based on RGB-assisted ICP, leveraging the advantages of RGB calibration to achieve long-range registration. We establish an experimental platform in a laboratory setting to verify the feasibility of the algorithm, calculating the distances between points and the reconstructed plane through plane parameter fitting. The proposed method is further applied to actual airport pallet measurement scenarios, achieving a reconstruction accuracy of 3.6 cm within a working range of 8 m × 8 m × 8 m. 收起