Abstract: To address the issue of low detection accuracy in underwater target detection caused by color distortion and target occlusion, an underwater object detection algorithm named MFTC-Net, based on multimodal feature fusion and temporal context awareness, is proposed. First, a multimodal feature fusion module is constructed, which achieves adaptive fusion of optical image features and simulated sonar features through deformable convolutions and attention mechanisms, effectively enhancing the model's perception of blurred targets. Second, a temporal context-aware module is designed, utilizing optical flow estimation and a Gated Recurrent Unit (GRU) to capture motion information across consecutive frames, significantly improving the recognition accuracy for occluded targets. Finally, a context-decoupled detection head is introduced to optimize feature representation for classification and localization tasks. Experimental results on the RUOD dataset demonstrate that at an IoU threshold of 0.5, the mean Average Precision (mAP) and mean Recall (mRecall) are improved by 4.1% and 3.8%, respectively; within the IoU threshold range of 0.5:0.95, improvements of 4.5% and 4.3% are achieved, respectively. The proposed algorithm provides an effective solution for target detection in complex underwater environments.