Abstract
Partial registration for point clouds plays an important role in various fields such as 3D mapping reconstruction, remote sensing, unmanned driving, and cultural heritage protection. Unfortunately, partial registration is challenging due to difficulties such as the low overlap ratio of two point clouds and the perturbation in the orderless and sparse 3D point clouds. Thus, a variety of the 3D shape context descriptors are introduced for finding the optimal matching. However, extracting geometric features and descriptors are time consuming and easily degenerated by noise. To overcome these problems, we introduce a parallel coarse-to-fine partial registration method. Our contributions can be summarized as: Firstly, a robust coarse trimmed method is proposed to estimate the coarse overlap area and the initial transformation via fast bilateral denoising and parallel point feature histogram (PPFH) descriptor aligning. Secondly, an accelerated fine registration procedure is conducted by a parallel trimmed iterative closest point (PTrICP) method. Moreover, most parts of our coarse-to-fine workflow are accelerated under the Graphics Processing Unit (GPU) parallel execution mode for efficiency. Thirdly, we extend our method from the rigid registration to the isotropic scaling registration, which improves its applicability. Experiments have demonstrated that our method is feasible and robust in various situations, including the low overlap ratio, outlier, noise and scaling.
Highlights
Point cloud data processing is a hot topic in recent years [1], [2]
A parallel bilateral denoising method and parallel point feature histogram (PPFH) aligning between key point pairs are used to eliminate the majority of non-overlap information
Comparing with the nearest neighbor search (NNS) based parallel Iterative Closest Point (ICP) method [49], [50], our contributions in parallel registration lie in: 1) we provide Graphics Processing Unit (GPU) parallelization for all modules of the point cloud registration; 2) we implement the parallel trimmed ICP algorithm, which improves the robustness by adding the parallel overlap ratio estimation in each iteration; and 3) we use the high-level Thrust library in our parallel implementation, which strengthens the reliability of our algorithm
Summary
Point cloud data processing is a hot topic in recent years [1], [2]. Point cloud registration, in particular, is a fundamental problem in computer vision and remote sensing [3]–[10]. These techniques adjust the location of each point via different projection strategies to filter point cloud, and determine the filtered position using similarity measures between a point and its neighbors These kinds of noise smoothing algorithms cannot solve the low overlap ratio problem caused by missing points and outliers. Besides the aforementioned randomized aligning methods, the principal component analysis (PCA) [38] is often used to obtain several principal axes of two point sets These global matching methods align the centers and their principal axes for the coarse transformation, while they fail if these two point clouds are under the low overlap ratio. A coarse-to-fine GICP algorithm combines the plane-to-plane and point-to-point trimmed ICP, which balances the stability and accuracy by changing the neighborhood search range from wide-base to narrow-base in each iteration [37] This adaptive strategy still fails under the low overlap ratio.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
