Calibration of a rotating 2D LRF in unprepared environments by minimizing redundant measurement errors

This paper proposes a calibration method for a 3D laser scanning system built with a rotating 2D laser rangefinder (LRF) in unprepared environments. Although the combination of a 2D LRF with a rotating unit yields a low-cost and compact 3D scanner, construction misalignments must be calibrated to reduce measurement error. Existing calibration techniques are generally used either with a special target object or in prepared environments with flat surfaces, and this imposes a limitation on the applicable domain of the system. To solve this problem, we propose a novel calibration method using redundant measurements of 3D point clouds. The main idea of the proposed method is to focus on the fact that rotating LRF systems observe the same object twice during a single rotation at different times. Therefore, by minimizing the extent of misalignment of these redundant measurements, we can estimate the calibration parameters without using any special target object. We formulate this calibration by redundant measurements as a non-linear cost minimization problem. The results of the simulation and real-world experiments in challenging, real environments show that the proposed method works well in unprepared and even cluttered environments.