In recent years, the demand for 3D city model development has grown, as demonstrated by initiatives such as Project PLATEAU. In the creation of LOD2 building models, which are an essential component of 3D city models, the reconstruction of 3D roof models still heavily depends on manual work. To enhance productivity through automation, this study proposes a novel method for automatically generating high-precision 3D roof models using orthophotos and Digital Surface Models (DSMs) derived from aerial imagery. In the proposed method, a deep learning model is first applied to orthophotos and DSMs to extract 2D roof lines. Then, the extracted 2D roof lines are refined and polygonised to generate 2D roof models. Finally, planar fitting is performed on the point clouds generated from the DSMs within each 2D roof surface to generate 3D roof models. The results of experiments showed that the recall of 3D roof surface reconstruction was 0.395, and increased to 0.451 for 3D roof surfaces larger than 8 m².