Event
TLP: A Single-Parameter Robust Point Cloud Registration Method
1 reports · 1 sources
Seong Hun Lee, Patrick Vandewalle, and Javier Civera propose the TLP point cloud registration method: it uses only a truncated Lp function to score correspondences, filter hypotheses, and refine the model, requiring only one tunable parameter throughout. The authors claim that compared with multi-parameter methods that require tuning multiple parameters for each dataset, this design is more practical; by evaluating hypotheses in parallel on a GPU, registration with thousands of correspondences can on average be completed within 1 second. On public benchmark datasets, the authors report that TLP achieves higher accuracy than existing single-parameter methods and outperforms two commonly used multi-parameter methods on three of the four evaluation datasets, while maintaining competitive runtime. These comparison results are all claims made by the authors based on their evaluation.
Report timeline
- TLP: Single-Parameter Robust Point Cloud Registration with Truncated Lp Functions
The authors propose TLP, a point cloud registration method with only one tunable parameter. It uses truncated Lp functions with different p values to assign priority scores to correspondences, select hypotheses, and refine the final model. Targeting the high outlier rates in real-world 3D feature matching, the method uses GPU parallelism to evaluate hypotheses and can on average complete registration from thousands of correspondences within 1 second. On public benchmark datasets, TLP is more accurate than existing single-parameter methods and outperforms two commonly used multi-parameter methods on three of the four evaluated datasets, while maintaining competitive runtime.
arXiv cs.RO ·