Hi^2-GSLoc: Dual-Hierarchical Gaussian-Specific Visual Relocalization for Remote Sensing

2025-07-21

AI Summary

To address low visual relocalization accuracy, high computational overhead, and poor scalability in remote sensing and UAV scenarios, this paper proposes a dual-level Gaussian-specific matching framework. We introduce 3D Gaussian lattices as a lightweight dense scene representation and design a sparse-dense two-stage 6-DoF pose estimation architecture. The method integrates partitioned training, GPU-accelerated parallel matching, consistency-aware rendering-based sampling, landmark-guided detection, dense rasterization-based matching, and geometric-semantic joint verification. Extensive evaluation on synthetic data, public benchmarks, and real-world UAV flight experiments demonstrates significant improvements in localization accuracy and recall, alongside reduced memory footprint and inference latency, while effectively suppressing outlier poses. This work achieves a superior trade-off among accuracy, efficiency, and scalability, establishing a novel paradigm for large-scale remote sensing relocalization.

Technology Category

• Intelligent Robots: Localization, Mapping, and Navigation
• Computer Vision: Remote Sensing / Geospatial AI
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Security and Privacy: Large-scale security measurements

Abstract

Visual relocalization, which estimates the 6-degree-of-freedom (6-DoF) camera pose from query images, is fundamental to remote sensing and UAV applications. Existing methods face inherent trade-offs: image-based retrieval and pose regression approaches lack precision, while structure-based methods that register queries to Structure-from-Motion (SfM) models suffer from computational complexity and limited scalability. These challenges are particularly pronounced in remote sensing scenarios due to large-scale scenes, high altitude variations, and domain gaps of existing visual priors. To overcome these limitations, we leverage 3D Gaussian Splatting (3DGS) as a novel scene representation that compactly encodes both 3D geometry and appearance. We introduce $mathrm{Hi}^2$-GSLoc, a dual-hierarchical relocalization framework that follows a sparse-to-dense and coarse-to-fine paradigm, fully exploiting the rich semantic information and geometric constraints inherent in Gaussian primitives. To handle large-scale remote sensing scenarios, we incorporate partitioned Gaussian training, GPU-accelerated parallel matching, and dynamic memory management strategies. Our approach consists of two stages: (1) a sparse stage featuring a Gaussian-specific consistent render-aware sampling strategy and landmark-guided detector for robust and accurate initial pose estimation, and (2) a dense stage that iteratively refines poses through coarse-to-fine dense rasterization matching while incorporating reliability verification. Through comprehensive evaluation on simulation data, public datasets, and real flight experiments, we demonstrate that our method delivers competitive localization accuracy, recall rate, and computational efficiency while effectively filtering unreliable pose estimates. The results confirm the effectiveness of our approach for practical remote sensing applications.

Problem

• Improving visual relocalization accuracy in remote sensing
• Reducing computational complexity in large-scale scenes
• Addressing domain gaps in existing visual priors

Innovation

• Uses 3D Gaussian Splatting for scene representation
• Dual-hierarchical sparse-to-dense relocalization framework
• GPU-accelerated parallel matching for large-scale scenes