As an important form of historical and cultural heritage, the timely detection and assessment of surface disease on ancient city walls are of great significance for cultural relic conservation. Traditional manual inspection methods are inefficient, costly, and highly subjective, which make them unsuitable for large-scale surveys of wall surface disease. To address this issue, this study proposes an automated method for detecting surface disease on ancient city walls by integrating three-dimensional reconstruction with deep learning. First, a 3D model of the ancient city wall is obtained using oblique photogrammetry, and the real appearance of the wall surface is restored through texture mapping. Then, an intelligent viewpoint planning strategy is employed based on the spatial structural characteristics of the wall to generate a virtual camera traversal path that fully covers the wall surface. Next, the YOLOv10 object detection algorithm is used to identify multiple types of deterioration, such as weathering and mortar joint loss, in the rendered images from each viewpoint. Finally, depth information and camera parameters are used to map the two-dimensional detection results onto the 3D model through geometric transformation, enabling accurate spatial localization and visual annotation of surface defects. Experimental results demonstrate that the proposed method efficiently and accurately achieves fully automatic detection of wall surface disease. The detection results are intuitively visualized as colored patches on the 3D model and can be used to generate detailed statistical reports on defect distribution. This research provides a new technical approach for the digital conservation and intelligent monitoring of ancient architectural heritage, offering promising application prospects.