Aerial-ground LiDAR place recognition with patch-level self-supervised learning and expanded reciprocal re-ranking

2026-06-16

AI Summary

This work addresses the challenge of cross-view place recognition between aerial and ground LiDAR point clouds, which is hindered by viewpoint discrepancies and domain shift, as well as high false-match rates in initial retrieval. To tackle this, the paper proposes a retrieval network that integrates multi-scale block-level self-supervised learning with scene-level feature extraction, introducing block-level self-supervised learning to this task for the first time. Furthermore, an Expansion Reciprocal (ER) re-ranking algorithm is devised, which leverages the spatial structure and neighborhood semantic consistency of ALS point clouds to refine the similarity matrix without requiring additional training. On the CS-Urban-Scenes dataset, the method improves Recall@1 by 9.8% and Recall@1% by 3.2%; ER re-ranking further boosts Recall@1 by 10.2% on CS-Urban-Scenes and by 4.9% on CS-Campus3D.

Abstract

LiDAR place recognition determines one's position on a prior point cloud map. The most studied ground-level LiDAR place recognition suffers from pre-visit requirements, incomplete coverage, and limited perspectives. Using pre-acquired, full-coverage Airborne Laser Scanning (ALS) data as an aerial prior map overcomes these drawbacks, making cross-view place recognition necessary and advantageous. However, aerial-ground LiDAR place recognition faces significant challenges, including the domain gap between aerial and ground point clouds, and false positives during initial retrieval. To address these challenges, we present a novel retrieval and re-ranking framework for aerial-ground LiDAR place recognition. Based on the priors that neighboring point cloud patches share similar semantics with anchor patch, our retrieval network introduces patch-level self-supervised learning modules at multiple scales and integrates with scene-level learning to improve global feature discriminativeness between aerial and ground point clouds. Furthermore, leveraging the structured spatial distribution of ALS point clouds, we introduce an Expanded Reciprocal (ER) re-ranking algorithm to exploit neighborhood information maximally and refine each feature based on neighbor features, which are then used to update the similarity matrix for final ranking. Extensive experiments demonstrate that our retrieval network outperforms existing state-of-the-art (SOTA) methods, achieving a 9.8\% improvement in average Recall@1 and a 3.2\% improvement in average Recall@1\% on the CS-Urban-Scenes, while also showing the best performance on the CS-Campus3D dataset. Additionally, our ER re-ranking algorithm further boosts the average Recall@1 by 4.9\% on CS-Campus3D and 10.2\% on CS-Urban-Scenes without additional training.

Problem

• aerial-ground LiDAR
• place recognition
• cross-view
• domain gap
• false positives

Innovation

• patch-level self-supervised learning
• expanded reciprocal re-ranking
• aerial-ground LiDAR place recognition
• cross-view retrieval
• point cloud semantics