<p>Sonar imaging plays a pivotal role in the detection of targets underwater. However, the performance of most sonar detection methods is suboptimal due to the complexity of the underwater environment and its susceptibility to noise interference. To address these issues, this paper proposes a SCR-YOLOv8 algorithm based on an enhanced YOLOv8 architecture. This algorithm aims to overcome the challenges associated with traditional sonar target detection. The proposed approach involves two primary modifications: first, the Conv module in the trunk and neck networks is replaced with the more efficient SPDConv. Second, the CCFM module is introduced to reduce the model size and number of parameters. Subsequently, the Spatial Channel Reconstruction Module (SCRM) is employed. The design of the feature extraction and fusion stage of the model is intended to enhance the model’s efficiency in extracting contextual information from both spatial and channel dimensions. Finally, the Inner-CIoU is employed in lieu of the CIoU to achieve regression results that are both faster and more effective. The experimental results demonstrate that, in comparison with the baseline model, SCR-YOLOv8 enhances precision, recall, and mAP50 by 2.9%, 5.8%, and 2.3%, respectively. Concurrently, the model size and the computational complexity are diminished by 40.2% and 21.4%, respectively.Moreover, a frame rate of 91 FPS is attained, meeting the criteria for real-time detection.</p>