Abstract: Infrared weak small target detection in complex backgrounds remains challenging in the field of computer vision owing to issues such as an extremely small target scale， a low signal-to-noise ratio， and strong background noise. Most existing detection methods are primarily designed for general targets and do not fully consider the specific characteristics of weak small targets， nor can they effectively fuse cross-modal information. To address these issues， this study proposed a weak small target detection framework based on electro-optical and radar dual-modal fusion， with the aim of overcoming the limitations of existing methods and enhancing target distinguishability and localization accuracy. First， a detailed and semantic space-frequency fusion module was designed， which combined spatial， frequency， and semantic information to enhance the feature representation capability of weak small targets. Second， a dynamic encoding-decoding feature fusion module was introduced to enhance the interaction and fusion of multi-scale features. In addition， a modality-specific hybrid attention module was employed to strengthen the contextual modeling of weak small target regions. To address the issue of small target regression instability， a weak small target-aware loss function was proposed to improve the stability of predicted bounding boxes. Finally， a hybrid expert feature alignment decision system was designed to achieve effective collaborative fusion of infrared and radar information， thereby enhancing the robustness of the overall detection. Experimental results showed that the proposed method outperformed existing methods in terms of detection performance on the IRSTD-1K and NUDT-SIRST datasets， validating the effectiveness and advantages of the framework for weak small target detection.