A Target-Free Joint Calibration Method for LiDAR Pose and Internal Parameters

The LiDAR rotary scanning system faces the accuracy challenge caused by installation errors in 3D reconstruction applications. The traditional method based on calibration objects has problems such as cumbersome scene construction and low efficiency. Moreover, calibrating the axis parameters and internal parameters separately will lead to the accumulation of errors. This paper proposes a target-free joint calibration method. By using structured features such as architectural plans, multiple target planes (such as walls, floors, etc.) are manually selected. The planes are required not to be parallel. In the experiment, the included angles of the planes can generally be taken as 75°-105°, and multi-frame point clouds are converted to a unified coordinate system based on the rotation angle information recorded by the servo motor. Theoretically, for the same target plane, the point clouds of each frame should be highly coincident. Based on this principle, this paper designs a calibration algorithm based on minimizing the thickness of the plane. It uses an iterative optimization algorithm to simultaneously adjust the internal parameters of the axis participants to minimize the thickness of the point cloud on the target plane, thereby significantly improving the calibration accuracy. The experimental results show that compared with the calibration method only for axis parameters, the accuracy of this method has increased by 3.8%.