Fast Point Feature Histograms (FPFH) for 3D Registration
In our recent work [1], [2], we proposed Point Feature Histograms (PFH) as robust multi-dimensional features which describe the local geometry around a point p for 3D point cloud datasets. In this paper, we modify their mathematical expressions and perform a rigorous analysis on their robustness and complexity for the problem of 3D registration for overlapping point cloud views. More concretely, we present several optimizations that reduce their computation times drastically by either caching previously computed values or by revising their theoretical formulations. The latter results in a new type of local features, called Fast Point Feature Histograms (FPFH), which retain most of the discriminative power of the PFH. Moreover, we propose an algorithm for the online computation of FPFH features for realtime applications. To validate our results we demonstrate their efficiency for 3D registration and propose a new sample consensus based method for bringing two datasets into the convergence basin of a local non-linear optimizer: SAC-IA (SAmple Consensus Initial Alignment).
PDF AbstractCode
Tasks
Results from the Paper
Task | Dataset | Model | Metric Name | Metric Value | Global Rank | Benchmark |
---|---|---|---|---|---|---|
Point Cloud Registration | 3DMatch (at least 30% overlapped - FCGF setting) | RANSAC-2M | Recall (0.3m, 15 degrees) | 66.1 | # 7 | |
Point Cloud Registration | 3DMatch Benchmark | FPFH + RANSAC | Feature Matching Recall | 44.2 | # 15 | |
Point Cloud Registration | 3DMatch (trained on KITTI) | FPFH | Recall | 0.136 | # 5 | |
Point Cloud Registration | ETH (trained on 3DMatch) | FPFH | Feature Matching Recall | 0.221 | # 8 | |
Recall (30cm, 5 degrees) | 66.34 | # 9 | ||||
Point Cloud Registration | FPv1 | FPFH-8M | Recall (3cm, 10 degrees) | 9.51 | # 8 | |
RRE (degrees) | 4.347 | # 2 | ||||
RTE (cm) | 1.900 | # 2 | ||||
Point Cloud Registration | KITTI (FCGF setting) | RANSAC | Recall (0.6m, 5 degrees) | 34.2 | # 10 |