🤖 AI Summary
Existing underwater sonar loop closure detection methods rely on 2D projection or handcrafted keypoint extraction, leading to geometric information loss; learning-based or bag-of-words approaches require extensive preprocessing (e.g., model training or vocabulary construction), suffering from poor generalizability. This paper proposes a projection-free, keypoint-free, and pretraining-free 3D point cloud loop closure detection method operating directly on raw sonar point clouds. We construct per-point rotation-invariant structural feature maps by jointly encoding geometric structure, surface normals, and curvature, and introduce a novel structural similarity metric. The method significantly enhances robustness in feature-sparse environments and cross-scene adaptability. Evaluated on real-world Seaward datasets from Antarctic deep-sea and inland river/lake environments, it outperforms state-of-the-art keypoint-based and learning-based methods in accuracy, while eliminating all preprocessing requirements.
📝 Abstract
In this letter, we propose a point cloud structural similarity-based loop detection method for underwater Simultaneous Localization and Mapping using sonar sensors. Existing sonar-based loop detection approaches often rely on 2D projection and keypoint extraction, which can lead to data loss and poor performance in feature-scarce environments. Additionally, methods based on neural networks or Bag-of-Words require extensive preprocessing, such as model training or vocabulary creation, reducing adaptability to new environments. To address these challenges, our method directly utilizes 3D sonar point clouds without projection and computes point-wise structural feature maps based on geometry, normals, and curvature. By leveraging rotation-invariant similarity comparisons, the proposed approach eliminates the need for keypoint detection and ensures robust loop detection across diverse underwater terrains. We validate our method using two real-world datasets: the Antarctica dataset obtained from deep underwater and the Seaward dataset collected from rivers and lakes. Experimental results show that our method achieves the highest loop detection performance compared to existing keypoint-based and learning-based approaches while requiring no additional training or preprocessing.