Aiming at the problem that a large number of existing artificial intelligence algorithms are difficult to detect stably due to the complex attitude and variable scale of marine targets, a detection algorithm based on depth supervision and improved YOLOv8 is proposed. Firstly, a multi-scale convolution module is designed to extract the feature information of the target's multi-receptive fields and reduce the missed detection rate. Then, a deep supervision network is added to improve the utilization ratio of deep class information and shallow location information, and optimize the performance of target feature extraction of the backbone network. Finally, a channel attention mechanism is introduced in the detection head to filter the irrelevant information and enhance the recognition rate of the key features. The experimental results in the marine target dataset show that the mAP value of the improved algorithm reaches 93.69% and the recall rate reaches 85.16%, which is 7.38 and 8.52 percentage points higher than the original model respectively, and is better than the classical algorithm and the novel algorithm, The detection time is about 14ms, which can meet the requirements of real-time target detection at marine and provide effective technical references for shipping management and marine accident prevention.