RGB-SAR image fusion detection can fully utilize complementary information to achieve better performance than unimodal detection, which has attracted increasing attention. However, it still faces challenges like large inter-modal differences and difficulties in detecting small targets. To address these issues, this paper proposes an improved cross-modal fusion detection network based on YOLOv11. First, the single-branch backbone network is replaced with a dual-branch structure, and the SPDConv module is introduced to reduce information loss during downsampling. Second, a feature fusion module based on a content-guided attention mechanism is introduced to enable effective cross-modal feature fusion through adaptive weight assignment. Additionally, an extra small-target detection head is incorporated to enhance detection capability by combining shallow detail features with high-level semantic information. Experiments on the VEDAI-VIPSAR dataset demonstrate that the proposed method effectively overcomes the limitations of single-sensor detection and achieves superior performance in complex scenarios.