【Abstract】 The recognition of capillary non-perfusion areas in fundus fluorescein angiography(FFA) images is crucial for the early diagnosis and treatment of diabetic retinopathy. However, due to interference from retinal exudates, the accurate detection of CNP in fundus fluorography images using the traditional LadderNet model is limited. In order to address these issues and achieve intelligent non-perfusion areas segmentation in fundus vessels while reducing the labor cost associated with manual interpretation by doctors, an intelligent segmentation model for fundus fluorography based on Enhance LadderNet is proposed. Firstly, the convolution module in the traditional LadderNet is replaced by the Vgg module. This modification enhances the feature extraction ability of the convolution layer through parameter sharing and sparse interaction between convolutional layers, thereby improving network generalization and efficiency. Secondly, an attention mechanism is incorporated into the intelligent model. This mechanism enables the model to focus more on important features and regions within the images, reducing overfitting and enhancing feature expression. Finally, ablation experiments using real data from a collaborative hospital is conducted, to evaluate the performance of the Enhanced LadderNet model compared to the traditional LadderNet and other models. The results of these experiments demonstrate the proposed model significantly improvs the accuracy of segmenting capillary non-perfusin areas.