Abstract: Polarimetric synthetic aperture radar (PolSAR) can acquire more comprehensive polarization information of targets compared to single- or dual-polarization SAR (synthetic aperture radar, SAR). However, due to the relative scarcity of fully polarimetric SAR ship detection datasets, deep learning-based research on fully polarimetric SAR target detection remains limited. This paper investigates and analyzes typical polarimetric decomposition parameters and proposes a ship detection method based on an improved YOLO11n network. In this method, the dominant scattering features derived from Cloude decomposition are synthesized into pseudo-color images as network inputs to enhance detection performance. The proposed approach replaces the original CIoU (Complete-IoU) loss with Shape-IoU in the detection network, incorporates a Multidimensional Collaborative Attention (MCA) mechanism in the neck network, and introduces the InceptionNeXt-MCA module to balance detection accuracy and training efficiency. Experiments conducted on the publicly available 2024 FPSD (Fully Polarized Ship Detection) dataset demonstrate that the proposed method effectively improves ship target detection accuracy, achieving an AP50 of 94.2%, precision of 91.3%, and recall of 92.7%.