Abstract: In order to solve the problems of inaccurate target positioning, false detection and missed detection caused by large target scale difference and complex background interference in the process of infrared ship detection, an improved infrared ship detection algorithm of YOLOv11n was proposed. Firstly, the C3k2_PMFEM module was designed as the feature extraction module of the backbone network to enhance the multi-scale feature expression ability of the model. Secondly, the C2PSA module was improved to C2EMA by using the efficient multi-scale attention mechanism EMA, and the attention to the target region in the complex background was improved by capturing the long-distance dependence of the channel and spatial dimensions. Thirdly, based on the detail enhanced convolution, a shared detail enhanced convolutional detection head SDCD was proposed to strengthen the interaction of feature information at different scales, improve the perception of image edge information, and reduce parameter redundancy. Finally, Wise-ShapeIoU is used as the bounding box loss function to reduce the negative gradient caused by low-quality samples and improve the regression accuracy of the bounding box. Experimental results on the infrared ship dataset show that the improved algorithm is 2.5% and 2.1% higher than the original algorithm on mAP50 and mAP50-95, respectively, and the number of parameters is reduced by 12.4% compared with the original algorithm, which achieves a good balance between detection accuracy and parameter quantity.