3D reconstruction from single view is a very difficult problem in computer vision and computer graphics. Much different from method using presently, we proposed a feasible and efficient method of 3D building reconstruction method by using non-complete topological analysis in this paper, which rapidly achieves the process of recognizing architectures based one or two images. Firstly, the outline of visible parts of architectures is extracted from image and is analyzed to get the incomplete topology of architectures. Next, the resulted topology will be divided into different primitive geometries, and these geometries are used to match models in databases finally. Experiments show that our method VS SSDA has better effectiveness.