ARAS-Net: A Meta-Learning-Based Few-Shot Small-Target Detection Network for Aerial Infrared Remote Sensing Images

Recent advances in infrared small-target detection and few-shot learning have yielded promising outcomes; however, the existing approaches address these challenges separately, resulting in limited generalization and robustness in complex aerial remote sensing environments. To overcome this limitation, we propose ARAS-Net, a meta-learning-based few-shot small-target detection framework tailored for aerial infrared remote sensing images. The proposed method incorporates an attention-guided context mechanism that enhances the distinction between targets and background in cluttered scenes, a mutual aggregation strategy that promotes rapid adaptation and generalization under limited sample conditions, and a multiscale feature enhancement design that effectively captures and represents small targets. Comprehensive experiments on benchmark aerial infrared datasets confirm that our approach consistently surpasses state-of-the-art methods in both detection accuracy and robustness under few-shot settings. The source code is available at https://github.com/jlm-138/ARAS-Net