Research map / Point clouds / Registration and matching
RKHS-BA: A Robust Correspondence-Free Multi-View Bundle Adjustment Framework for Semantic Point Clouds
Ray Zhang, Jingwei Song, Xiang Gao, Junzhe Wu, Tiany Liu, Jinyuan Zhang, Ryan M. Eustice, Maani Ghaffari
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add semantic labels to color RKHS-BA | RKHS-BA · TartanAir sliding-window BA | mean ATE (m): 0.6440 -> 0.5838 | Improved |
| replace SemanticCVO frame-to-frame with multi-frame RKHS-BA backend | Semantic RKHS · TartanAir sliding-window BA | mean ATE (m): 1.9022 -> 0.5838 | Improved |
| add semantic labels to intensity RKHS-BA | RKHS-BA · SemanticKITTI loop sequences | mean translational drift (%): 0.6180 -> 0.6160 | Improved |
| add RKHS-BA refinement on top of MULLS pose graph | MULLS PGO · SemanticKITTI loop sequences | mean translational drift (%): 0.7226 -> 0.6160 | Improved |
| add RKHS-BA refinement on top of MULLS pose graph | MULLS PGO · SemanticKITTI loop sequences | mean rotational drift (deg/m): 0.0030 -> 0.0027 | Improved |
| use correspondence-free RKHS-BA instead of point-to-feature BALM | BA backend · SemanticKITTI loop sequences | mean rotational drift (deg/m): 0.0039 -> 0.0027 | Improved |
| use RKHS-BA instead of HBA | BA backend · SemanticKITTI loop sequences | mean translational drift (%): 0.7570 -> 0.6160 | Improved |
| increase color noise variance to 80 (colors in [0,255]) (ablation) | Semantic RKHS-BA · TartanAir four-frame BA | registration robustness | Hurt |
| increase label distribution noise to 0.2 (ablation) | Semantic RKHS-BA · TartanAir four-frame BA | registration robustness | Hurt |
| add semantic labels to RKHS-BA registration | RKHS-BA · TartanAir four-frame registration | running time | Improved |
| add semantic labels to color RKHS-BA | RKHS-BA · TartanAir four-frame registration | pose error | No clear effect |
| add proposed rotation initialization to CVO | CVO · Bunny two-view global registration (90/180 deg) | pose error | Improved |
Closest papers by meaning
- Generalized-CVO: Fast and Correspondence-Free Local Point Cloud Registration with Second Order Riemannian Optimization (CVPR 2026 · similarity 0.88)
- C-GenReg: Training-Free 3D Point Cloud Registration by Multi-View-Consistent Geometry-to-Image Generation with Probabilistic Modalities Fusion (CVPR 2026 · similarity 0.86)
- GraphI2P: Image-to-Point Cloud Registration with Exploring Pattern of Correspondence via Graph Learning (CVPR 2025 · similarity 0.86)
- Correspondence-Free SE(3) Point Cloud Registration in RKHS via Unsupervised Equivariant Learning (ECCV 2024 · similarity 0.86)
- GeoNLF: Geometry guided Pose-Free Neural LiDAR Fields (NeurIPS 2024 · similarity 0.85)
- Unlocking Generalization Power in LiDAR Point Cloud Registration (CVPR 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
