In response to the frequent occurrence of road surface diseases in port areas and the inadequacy of traditional detection methods to cope with large-scale and high-frequency road inspection tasks, as well as the inability to simultaneously locate and finely segment diseases, this paper proposes a road disease recognition method based on Mask R-CNN. Firstly, a large-scale road dataset containing 13005 high-definition images and corresponding pixel annotation information is constructed. ResNet-101 and FPN are adopted as the feature extraction modules, combined with RPN, RoI Align, and mask branch modules for rapid and accurate disease recognition and pixel-level segmentation. The model achieved an average precision of 87.4%, an average recall of 88.6%, and an average F1-Score of 0.880 in identifying four types of pavement distress: transverse cracks, longitudinal cracks, transverse repairs, and longitudinal repairs. Additionally, under the condition of an Intersection over Union (IoU) threshold of 0.5, the model"s mAP_50 reached 88.1%. In practical application tests on port area roads, the associated software demonstrated an mAP_50 of 85.4%, with clear boundaries generated for the masks. These results indicate that the proposed method effectively reduces the technical difficulty of distress detection and assessment, enhances the efficiency and objectivity of road maintenance evaluations, and improves the level of intelligence in traffic inspections of port area roads.