Xianhao Wu

Improving visual quality in unmanned aerial vehicle (UAV) remote sensing scenarios under adverse weather conditions presents a significant challenge, as different types of weather degradation exhibit distinct characteristics. Most existing methods are trained independently for specific types of degradation, which severely limits their adaptability to dynamic real-world environments. To tackle these challenges, we employ a multiple-in-one image restoration strategy for UAV remote sensing, aiming to address multiple weather-induced degradations through a unified training model. First, we construct a high-quality dataset called UAV-Weather6k, which includes adverse weather conditions such as rain, haze, and snow. Subsequently, we propose an effective scale-aware Trident Mamba for unified image restoration. The framework integrates a multi-resolution and a multi-patch parallel multi-branch architecture, where each branch employs different multi-scale learning to effectively capture features under diverse weather conditions. Extensive experiments demonstrate the effectiveness of our model, achieving competitive performance compared with state-of-the-art methods.