An Improved Method for Extrinsic Calibration of Tilting 2D LRF

This paper proposes an improved calibration method to accurately estimate extrinsic calibration parameters of a tilting 2D Laser Range Finder (LRF). Tilting 2D LRF (a low cost 3D scanner device) with a unidirectional rotating platform has been widely used in robotics applications to scan the 3D environment. Ideally, the tilt axis of rotating mechanism should pass through the optical rotation centre of the 2D LRF. However, due to misalignment during assembling of 2D LRF with the rotating platform, the centres of rotation may not coincide with each other. Though, the system must be calibrated to align both the centres of rotation so as to improve the accuracy in building 3D point cloud of the environment. Unlike the previous calibration techniques, the main advantage of the proposed method is that it accurately estimates all 6-DOF calibration parameters, especially rotation and translation calibration parameters along motor rotation axis between 2D LRF and rotating platform without any additional hardware, camera or rolling/bidirectional rotation mechanism. The proposed method utilizes the normal vector to the calibration board plane and coordinates of laser points at the endpoints of the extracted calibration board line in the 2D scan to obtain these remaining calibration parameters. The obtained parameters are then refined using Levenberg-Marquardt non-linear optimization algorithm. The performance of the algorithm is validated on a range of real as well as synthetic data and the estimated parameters with the proposed approach exhibits 24.54% reduction in RMS (root mean square) error as compared to the conventional approach. Furthermore, qualitative and quantitative analysis shows that the proposed method produces accurate results and is able to reduce the artifacts along the rotation axis in the 3D point cloud which usually appear in the point cloud obtained from the conventional approach.