Fine-Grained Cross-View Localization via Local Feature Matching and Monocular Depth Priors

2025-09-11

AI Summary

To address registration inaccuracies and information loss in cross-view localization caused by perspective distortion and severe height compression, this paper proposes a fine-grained, no-fine-tuning cross-view localization method. Our approach jointly optimizes local feature matching between ground-level and aerial images and monocular depth priors, directly lifting matched keypoints into bird’s-eye view (BEV) space. We further introduce scale-aware Procrustes alignment to estimate 3-degree-of-freedom camera pose. The method accommodates arbitrary metric or relative depth inputs and models cross-view correspondences via weakly supervised learning. Experiments demonstrate significant improvements in localization accuracy and robustness under challenging conditions—including cross-region generalization and unknown camera orientation—while maintaining high interpretability and practical deployability.

Technology Category

• Computer Vision: Learning & Optimization for CV
• Machine Learning: Multi-instance/Multi-view Learning
• Intelligent Robots: Localization, Mapping, and Navigation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

We propose an accurate and highly interpretable fine-grained cross-view localization method that estimates the 3 Degrees of Freedom pose of a ground-level image by matching its local features with a reference aerial image. Previous methods typically transform the ground image into a bird's-eye view (BEV) representation and then align it with the aerial image for localization. However, this transformation often leads to information loss due to perspective distortion or compression of height information, thereby degrading alignment quality with the aerial view. In contrast, our method directly establishes correspondences between ground and aerial images and lifts only the matched keypoints to BEV space using monocular depth prior. Notably, modern depth predictors can provide reliable metric depth when the test samples are similar to the training data. When the depth distribution differs, they still produce consistent relative depth, i.e., depth accurate up to an unknown scale. Our method supports both metric and relative depth. It employs a scale-aware Procrustes alignment to estimate the camera pose from the correspondences and optionally recover the scale when using relative depth. Experimental results demonstrate that, with only weak supervision on camera pose, our method learns accurate local feature correspondences and achieves superior localization performance under challenging conditions, such as cross-area generalization and unknown orientation. Moreover, our method is compatible with various relative depth models without requiring per-model finetuning. This flexibility, combined with strong localization performance, makes it well-suited for real-world deployment.

Problem

• Estimating ground image pose via local feature matching with aerial imagery
• Addressing information loss from perspective distortion in cross-view localization
• Supporting both metric and relative depth models for flexible deployment

Innovation

• Direct local feature matching between ground and aerial images
• Lifting matched keypoints using monocular depth priors
• Scale-aware Procrustes alignment for camera pose estimation