The UAV Image Matching System Based on Multi-Feature Fusion Representation
Xianfeng Song, Yi Zou, Zheng Shi, Yanfeng Yang, Dacheng Li · 2023
Image matching, a process that identifies and aligns pixel-level attributes shared or similar between two images, is a critical component in the perception and navigation capabilities of unmanned aerial vehicle (UAV) systems. Current research methodologies, which predominantly employ either feature-based or deep learning-based methods, present significant challenges. Feature-based methods often lack robustness, while deep learning-based methods necessitate the sacrifice of substantial specific samples and computational resources. To address these issues, we propose a novel UAV image matching system predicated on multi-feature fusion representation. The results indicate that our proposed system enhances robustness and augments overall matching accuracy by a factor of 2.5 to 6 when juxtaposed with existing methods. Code is made publicly available at https://github.com/songxf1024/DeFusion.