Comparing Different Strategies for Frame-to-Frame Rigid Registration of Point Clouds

Abstract

Pairwise surface rigid registration aims to find the rigid transformation that best register two surfaces represented by point clouds. This work presents a comparison between seven algorithms, with different strategies to tackle rigid registration tasks. We focus on the frame-to-frame problem by using both point clouds and a RGB-D video stream in the experimental results. The former, is considered under different viewpoints, with the addition of outliers and noise. Once the ground truth rotation is provided, we discuss four different metrics to measure the rotation error in this case. The video sequence with depth information is segmented to get the target object. Next, the registration algorithms are applied and the average root mean squared error is computed. Since the ground truth is not available in this case, we develop a superposition strategy to visually check performance of the algorithms. Besides, we analyse the robustness of the techniques against spatial and temporal sampling rates.