Improved Bag-of-Words Image Retrieval with Geometric Constraints for Ground Texture Localization

2025-05-16

AI Summary

Ground-texture-based localization from downward-looking UAV/robot vision poses challenges for global localization and SLAM loop closure detection due to the limited discriminability and geometric instability of conventional Bag-of-Words (BoW) image retrieval. Method: This paper proposes an Approximate k-Means (AKM)-based soft-assignment vocabulary construction method, incorporating ground-texture-specific geometric constraints—namely, constant scale and consistent orientation—for the first time. It further designs a dual-mode retrieval architecture balancing high accuracy and real-time performance. Contribution/Results: The proposed method significantly improves global localization accuracy and enhances both precision and recall in loop closure detection. Crucially, it retains full plug-and-play compatibility with existing BoW-based pipelines—requiring no hardware modifications or environmental adaptations—enabling seamless replacement of conventional BoW modules while delivering superior robustness and reliability.

Technology Category

• Computer Vision: Image and Video Retrieval
• Intelligent Robots: Localization, Mapping, and Navigation
• Search and Optimization: Local Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Ground texture localization using a downward-facing camera offers a low-cost, high-precision localization solution that is robust to dynamic environments and requires no environmental modification. We present a significantly improved bag-of-words (BoW) image retrieval system for ground texture localization, achieving substantially higher accuracy for global localization and higher precision and recall for loop closure detection in SLAM. Our approach leverages an approximate $k$-means (AKM) vocabulary with soft assignment, and exploits the consistent orientation and constant scale constraints inherent to ground texture localization. Identifying the different needs of global localization vs. loop closure detection for SLAM, we present both high-accuracy and high-speed versions of our algorithm. We test the effect of each of our proposed improvements through an ablation study and demonstrate our method's effectiveness for both global localization and loop closure detection. With numerous ground texture localization systems already using BoW, our method can readily replace other generic BoW systems in their pipeline and immediately improve their results.

Problem

• Enhancing ground texture localization accuracy using improved BoW
• Optimizing loop closure detection in SLAM with geometric constraints
• Providing high-accuracy and high-speed algorithm versions for different needs

Innovation

• AKM vocabulary with soft assignment
• Geometric constraints for orientation and scale
• High-accuracy and high-speed algorithm versions