Low-light Unmanned Aerial Vehicle (UAV) image enhancement is crucial for downstream tasks such as object detection and navigation. However, low-light UAV images often have complex illumination patterns and high noise levels. Existing mainstream low-light enhancement methods tend to introduce color cast and local overexposure on such images. To address these issues, we propose a Low-light UAV image enhancement network named UAV-LIENet, and we train it with three progressive sub-networks. UAV-LIENet first applies a Non-uniform Luminance Estimation Network (NLEN) to reconstruct a smooth and uniform illumination component. NLEN adopts quantile-clipping normalization and a parallel coarse-and-fine architecture for illumination estimation. Then, UAV-LIENet performs adaptive denoising and color restoration under the guidance of the estimated luminance component. For accurate and stable color restoration, we design an illumination-guided saturation constraint loss, which adaptively constrains saturation in the HSV space to reduce color cast and suppress oversaturation. To evaluate our method systematically, we build a low-light enhancement dataset named UAV-LLIE based on high-fidelity game-engine rendering. UAV-LLIE contains 6 typical aerial scenarios and 6,000 pixel-aligned image pairs. Experiments show that UAV-LIENet outperforms representative existing methods in both quantitative metrics and visual quality.
UAV (Unmanned Aerial Vehicle)-to-satellite self-localization is commonly treated as satellite tile retrieval. This makes large-area search tractable, but it also forces a continuous localization problem into a discrete ranking form. Once the task is defined this way, training naturally relies on hard positive–negative tile labels, and inference tends to read coordinates from the top-ranked tile center. The model, therefore, learns image identity more than geographic continuity, while the final estimate remains vulnerable to tile-center quantization and top-1 retrieval errors. We propose PFIRNet (Probability Field Inference Network), a continuous geographic posterior inference framework that reformulates retrieval outputs as evidence for coordinate estimation rather than discrete tile selection. It uses distance-aware geographic supervision to shape candidate responses according to metric proximity, lifts top-k candidates into a coordinate-space probability field, and applies risk-calibrated multi-peak verification to update the estimate only when an alternative posterior peak is sufficiently supported. On DenseUAV, PFIRNet reduces the median localization error to 9.84 m and outperforms both one-stage retrieval methods and two-stage matching baselines. It also remains more robust under sparse and non-aligned satellite galleries.