In recent years, with the rapid development of computer technology, 3D reconstruction technology has gradually entered people’s lives. 3D reconstruction is one of the more popular research directions in the field of computer vision. Three-dimensional reconstruction mainly studies the process of how to obtain three-dimensional information based on single-view or multi-view reconstruction. With the growing demand for computers to automatically obtain three-dimensional information of the surrounding environment, the practical application requirements for three-dimensional reconstruction technology are also getting higher and higher. Therefore, how to quickly and efficiently reconstruct 3D models has become an important research topic in the field of computer vision. In view of this practical application problem, this paper starts from the perspective such as Structure from Motion,Clustering Multi-view Stereo,Patch-based Multi-View Stereo and other aspects, and use the learning-based method to match the sift feature points. In this paper, we mainly research on several key technologies in the 3D reconstruction process.