Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2508.17034 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916014515552256 |
|---|---|
| author | Li, Jiayi Yao, Yuxin Lu, Qiuhang Zhang, Juyong |
| author_facet | Li, Jiayi Yao, Yuxin Lu, Qiuhang Zhang, Juyong |
| contents | Noisy, partially overlapping data and the need for real-time processing pose major challenges for rigid registration. Considering that feature-based matching can handle large transformation differences but suffers from limited accuracy, while local geometry-based matching can achieve fine-grained local alignment but relies heavily on a good initial transformation, we propose a novel dual-space paradigm to fully leverage the strengths of both approaches. First, we introduce an efficient filtering mechanism consisting of a computationally lightweight one-point RANSAC algorithm and a subsequent refinement module to eliminate unreliable feature-based correspondences. Subsequently, we treat the filtered correspondences as anchor points, extract geometric proxies, and formulate an effective objective function with a tailored solver to estimate the transformation. Experiments verify our method's effectiveness, as demonstrated by a 32x CPU-time speedup over MAC on KITTI with comparable accuracy. Project page: https://ustc3dv.github.io/DualReg/. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_17034 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | DualReg: Dual-Space Filtering and Reinforcement for Rigid Registration Li, Jiayi Yao, Yuxin Lu, Qiuhang Zhang, Juyong Robotics Computer Vision and Pattern Recognition Noisy, partially overlapping data and the need for real-time processing pose major challenges for rigid registration. Considering that feature-based matching can handle large transformation differences but suffers from limited accuracy, while local geometry-based matching can achieve fine-grained local alignment but relies heavily on a good initial transformation, we propose a novel dual-space paradigm to fully leverage the strengths of both approaches. First, we introduce an efficient filtering mechanism consisting of a computationally lightweight one-point RANSAC algorithm and a subsequent refinement module to eliminate unreliable feature-based correspondences. Subsequently, we treat the filtered correspondences as anchor points, extract geometric proxies, and formulate an effective objective function with a tailored solver to estimate the transformation. Experiments verify our method's effectiveness, as demonstrated by a 32x CPU-time speedup over MAC on KITTI with comparable accuracy. Project page: https://ustc3dv.github.io/DualReg/. |
| title | DualReg: Dual-Space Filtering and Reinforcement for Rigid Registration |
| topic | Robotics Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2508.17034 |