Abstract With advantages of 3D representation, non-contact measurements and intensive sampling capability, it has been a research hotspot to detect the potential damage area of bridges with point cloud by TLS. However, TLS is commonly used to detect a potential damage area by comparing multi-temporal point cloud data, which limits the timeliness of bridge inspection. Therefore, aiming to accurately detect the potential damage areas of bridges with single-temporal point cloud, this paper proposes a normalized normal vector constrained coordinate transformation method. First, the distribution of sharp features is revealed in a single-temporal point cloud at potential damage areas, and a neighborhood growth method constrained by the normal distance is proposed to eliminate the sharp features in point cloud, which is prone to cause incorrect or missing curvature values from point cloud. Second, a normalized normal vector constrained coordinate transformation method is proposed to construct Gaussian curvature model, which can improve the accuracy of point cloud curvature and accurately detect the potential damage areas in bridges. At last, an evaluation criterion is proposed to quantify the bridge condition by integrating the characteristics of small-span concrete bridges in urban areas with actual damage data from the experimental bridges. The experimental results show that the proposed method can effectively detect the potential damage areas of the measured bridges.