Abstract: The coverage rate of polymetallic nodules is a key indicator for assessing the distribution density and resource potential of deep-sea mining areas, and it is of great significance for achieving rapid and accurate estimation. In response to the characteristics of irregular nodule shapes, large scale differences, and surface attachment interference in deep-sea environments, a coverage rate assessment method of polymetallic nodule based on an improved YOLOv11n-seg has been proposed, which systematically enhances model performance through three key aspects: feature extraction, multi-scale fusion, and loss optimization. This method firstly introduces a Deformable Attention Transformer（DAT）, combining the local adaptive capability of deformable convolution with the global modeling advantages of transformers to achieve dynamic perception and robust feature extraction of multi-scale and irregular nodules. Secondly, a High-leval Screening Feature Pyramid Network（HS-FPN） is constructed to enhance the efficiency of semantic information transfer through a cross-layer bidirectional fusion mechanism, while a hierarchical lightweight compression is used to reduce redundant features, and cross-scale residual connections further improve sensitivity to small nodule targets. Finally, the SlideLoss loss function is adopted to dynamically adjust the confidence threshold of difficult samples, effectively alleviating the imbalance issue between positive and negative samples and promoting the model’s convergence stability and segmentation consistency under complex samples. Experimental results show that the proposed model significantly outperforms the original model on multiple performance metrics, with an increase of 0.6% in mean average precision（mAP@0.5:0.95）, an improvement of 0.8% in segmentation accuracy, a reduction of 28.6% in model parameters, and a decrease of 8.7% in computational load, as well as a notable acceleration in image segmentation speed. This research provides a reliable technical pathway for the efficient and accurate estimation of deep-sea polymetallic nodule coverage, which has significant practical application value for advancing the exploration and development of deep-sea mineral resources.