Aerial-ground LiDAR place recognition with patch-level self-supervised learning and expanded reciprocal re-ranking
Researchers have developed a novel framework for aerial-ground LiDAR place recognition, addressing challenges like the domain gap and false positives. Their approach utilizes patch-level self-supervised learning to enhance feature discriminativeness between aerial and ground point clouds. Additionally, an Expanded Reciprocal (ER) re-ranking algorithm leverages neighborhood information to refine features and improve final rankings. Experiments show significant improvements in recall rates on benchmark datasets like CS-Urban-Scenes and CS-Campus3D. AI
IMPACT Enhances the accuracy and robustness of autonomous navigation systems relying on LiDAR data.