Struct-Loc: Confidence-Aware Structural Localization Via Hierarchical Point Cloud Registration
Csaba Máté Józsa, Attila Bóta, Krisztián Zsolt Varga, Ferenc Kovács
AI summary
Problem
Visual localization degrades under challenging conditions like variable lighting and repetitive textures, while single-image methods lack context and probabilistic approaches are computationally expensive. Existing point cloud registration techniques also ignore correspondence reliability, treating all matches uniformly.
Approach
The method formulates localization as a hierarchical point cloud registration problem, regressing confidence scores for coarse structural regions and propagating them to weight fine-level matches in a pose solver.
Key results
- Outperforms strong baselines in accuracy and runtime on the LaMAR benchmark
- Achieves 100× global map compression compared to COLMAP
- Enables near real-time inference through efficient encoders and map caching
- Demonstrates high recall and robustness under real-world visual ambiguities
Why it matters
Enables reliable, scalable indoor localization for robotics and AR fleets in complex environments where vision-only methods fail.
Abstract
Localization systems often rely heavily on visual information, which can degrade under challenging conditions such as variable lighting, dynamic objects, or repetitive textures. To enhance robustness beyond single-image methods, we model localization as a structural point cloud registration problem, leveraging motion continuity and geometric consistency over time. This formulation reduces sensitivity to transient occlusions and appearance changes, enabling the system to resolve ambiguities that single-image techniques often cannot. In this work, we introduce Struct-Loc, a localization frame- work that advances structural point cloud registration through confidence-aware hierarchical localization. By estimating the reliability of structural regions and incorporating it into the matching process, Struct-Loc generates robust descriptors tailored for pose estimation. To achieve near real-time performance, Struct-Loc combines efficient point convolutional encoders, a caching mechanism, and a hierarchical coarse-to-fine matching strategy that progressively narrows the search space. It consistently outperforms strong baselines in both accuracy and runtime, while achieving a 100× compression of the global map compared to COLMAP, significantly improving storage efficiency. We validate Struct- Loc on the LaMAR benchmark, demonstrating its effectiveness and robustness under real-world conditions.