Space-based infrared target detection, with its all-weather, all-day capability and wide-area search, is widely used in national defense, early warning, and civil aviation search and rescue. However, due to long-range detection and environmental noise, targets often have weak, textureless appearances, low contrast, and low signal-to-noise ratios, making detection challenging. Traditional detection methods struggle to effectively distinguish between the target and background in such complex environments, leading to frequent false alarms and missed detections. Therefore, improving the detection accuracy of small targets in complex environments has become a key focus of current research. To address this issue, this study proposes a novel dynamic weighted fusion network (DWF-Net), which significantly improves the accuracy and robustness of target detection in infrared images by introducing dynamic weighting mechanisms and multi-level feature fusion strategies. Specifically, DWF-Net designs a dynamic weighting mechanism based on target spatial distribution and feature responses, allowing the network to automatically adjust the weights of different regions and feature channels. This enables effective discrimination between targets and backgrounds, thereby enhancing the detection accuracy of small targets. Additionally, DWF-Net employs a multi-level feature fusion strategy, finely integrating features from different levels and scales to fully exploit multi-dimensional feature information. This allows the network to accurately recognize targets at multiple scales, further improving detection accuracy and robustness. Experimental results on various datasets demonstrate that DWF-Net exhibits excellent performance in solving the problem of small target detection in complex backgrounds, providing a more efficient and reliable solution for infrared target detection.
Li et al. (2025) studied this question.