Multi-weather Cross-view Geo-localization Using Denoising Diffusion Models

TL;DR

This paper presents MCGF, a framework that uses denoising diffusion models to improve cross-view geo-localization across diverse and unseen weather conditions by jointly restoring images and extracting features.

Contribution

The paper introduces MCGF, a novel multi-weather geo-localization framework that dynamically adapts to unseen weather conditions through joint image restoration and feature extraction.

Findings

MCGF achieves competitive geo-localization accuracy across varying weather conditions.

The joint optimization improves robustness against unseen extreme weather.

Experimental results on University160k-WX validate the effectiveness of MCGF.

Abstract

Cross-view geo-localization in GNSS-denied environments aims to determine an unknown location by matching drone-view images with the correct geo-tagged satellite-view images from a large gallery. Recent research shows that learning discriminative image representations under specific weather conditions can significantly enhance performance. However, the frequent occurrence of unseen extreme weather conditions hinders progress. This paper introduces MCGF, a Multi-weather Cross-view Geo-localization Framework designed to dynamically adapt to unseen weather conditions. MCGF establishes a joint optimization between image restoration and geo-localization using denoising diffusion models. For image restoration, MCGF incorporates a shared encoder and a lightweight restoration module to help the backbone eliminate weather-specific information. For geo-localization, MCGF uses EVA-02 as a backbone for feature extraction, with cross-entropy loss for training and cosine distance for testing. Extensive experiments on University160k-WX demonstrate that MCGF achieves competitive results for geo-localization in varying weather conditions.