Abstract: Objectives The YOLOv8 algorithm has been widely recognized in the field of object detection due to its efficient processing and superior detection performance. However, in ship target detection problems, it still suffers from insufficient robustness, suboptimal multi-scale feature extraction, and an excessive number of parameters. To overcome these issues, a lightweight SAP-YOLOv8 algorithm with enhanced features representation is proposed. Methods Incorporating spatial depthwise convolution and dilated convolution into standard convolution to construct the SDGD module, thereby enhancing the suppression capability against sea clutter interference and the extraction capability of multi-scale features. Introduce the AIFI module from RT-DETR to replace the SPPF module, thereby enhancing the algorithm's contextual modeling capability in complex sea conditions. To optimize computational efficiency, a C3k2_PCCA module, based on the partial convolution and coordinate attention mechanisms, is constructed to reduce parameters and complexity while improving lightweight performance and runtime efficiency in complex sea conditions. Results Experiments on the public HRSID dataset show that, with nearly unchanged model size, SAP-YOLOv8 improves precision, recall, and mean average precision by 1.5%, 0.7%, and 1.6%, respectively, compared to the original algorithm, and it outperforms other classic algorithms in detection performance. Conclusions The SAP-YOLOv8 algorithm exhibits higher detection accuracy and operational efficiency, while demonstrating stronger robustness and practical value in complex sea conditions.