Models that are built based on point cloud construction usually have uneven surfaces and jagged edges. Aiming at the problem of poor visual effect of these models， this paper proposes a three-dimensional （3D） model optimization method based on the three-dimensional line feature constraint of oblique photography. First， based on the overlapping area， the quality evaluation and screening of the line segments are carried out. Second， the three-dimensional line segments are optimally reconstructed based on the iteration method with variable weights. Finally， the plane fitting and the correction and optimization of the model surface are obtained based on the triangulation， and the 3D lines are used to correct and optimize the triangulation at the edge of the building. The method is compared with the Poisson surface reconstruction method. The results show that the proposed method effectively eliminates planar convex hull and edge sawtooth； the edge precision is optimized to within one pixel， and the plane precision is optimized to 0.5 pixels. The overall accuracy of the model is nearly double of that the Poisson surface reconstruction method. Compared with the existing surface reconstruction algorithms of based on point cloud features， the proposed method greatly improves the visual effect of the model， maintains the features of the 3D model plane and edge， and improves the edge and surface accuracy of the model.