UAV-based infrared target detection presents inherent challenges, including low signal-to-noise ratios, texture degradation, and severe scale variations. To address these issues, we propose DFS-YOLO, an approach based on dynamic feature collaboration and efficient state-space modeling. We introduce a Dynamic Range-Calibrated Area Attention (DRCAA) module in the backbone to stabilize feature activations under strong thermal clutter. Within the neck architecture, an Efficient Attentional Scale-Sequence Fusion (EASF) strategy reduces cross-scale semantic misalignment and ensures precise spatial coherence. Additionally, an EfficientViM-based state-space module captures global contextual dependencies while maintaining linear computational complexity. Finally, the Content-Guided Triple-Attention Fusion (CGTAFusion) module maximizes feature discriminability by calibrating fusion representations across the channel, spatial, and pixel dimensions. Extensive experiments on the HIT-UAV and IRSTD-1k benchmarks validate the efficacy of the DFS-YOLO framework. Compared to the baseline YOLOv12, DFS-YOLO’s performance has been significantly improved, increasing mAP@50 and mAP@50-95 by 10.16% and 7.55% on HIT-UAV, and by 1.84% and 3.18% on IRSTD-1k, respectively. These quantitative gains establish DFS-YOLO as a highly robust and state-of-the-art solution for complex infrared aerial surveillance.

Existing remote sensing object detection methods struggle with challenges such as complex background interference, variable object scales, and class imbalance due to a lack of coordinated internal optimization. This paper proposes AFDNet, a novel RSOD algorithm that establishes an internal collaborative evolution mechanism to systematically enhance the model’s feature perception and localization capabilities in complex scenes. AFDNet achieves this through three tightly coupled, co-evolving components: (1) A channel–spatial dual-sensing module that adaptively focuses on crucial features and suppresses background noise. (2) A dynamic bounding box optimization module that integrates distance-aware and scale-normalization strategies, significantly boosting localization accuracy and regression robustness for multi-scale objects. (3) A Gaussian adaptive activation unit that enhances the model’s nonlinear fitting capability for better detail extraction under weak conditions. Extensive experiments on two public datasets, RSOD and NWPU VHR-10, verify the excellent performance of AFDNet. AFDNet achieved a leading 95.16% mAP@50 on the RSOD dataset and an astonishing 96.52% mAP@50 on the NWPU VHR-10 dataset, which is significantly better than the mainstream detection models. This study verifies the effectiveness of introducing internal co-evolution mechanisms and provides a novel and reliable solution for high-precision remote sensing target detection.