Iterative Closest Point (ICP) Performance Comparison Using Different Types of Lidar for Indoor Localization and Mapping
Z-H. Ouyang, S-A. Cui, P-F. Zhang, S-F. Wang, X. Dai and Q. Xia
Lidars (laser scanners) are usually used to perceive the surrounding environment and establish the environment map, thereby realizing indoor mobile positioning. Point cloud data’s registration method plays an important role in estimating the three-dimensional motion of a rigid object. Meanwhile, different lidars have different performance parameters like the resolution, which is likely to lead to their different performance when conducting indoor positioning. In this paper, we compare and analyse registration results of datasets obtained by different lidars. In the beginning, the fast point feature histograms (FPFH) descriptor is introduced in feature extraction, and the sample consensus method is combined to conduct the coarse registration. Secondly, fine registration of a nearly-aligned initial pose is provided by the iterative closest point (ICP) algorithm. Finally, experiments are conducted to compare registration runtime and registration error. The experimental results illustrate that high resolution improves feature recognition and reduces registration error. Among those single-layer and multilayer lidars, the 32-layer one possessed a better balance between the registration accuracy and the time-consumption.
Keywords: Lidar, iterative closest point (ICP), algorithm, registration, coordinate transformation, convergence