Stereo vision is increasingly applied to various fields of life, and three-dimensional (3D) reconstruction is one of the hottest topics in stereovision. Because the active 3D system is costly and susceptible to illumination, and the reconstruction results have problems in accuracy and robustness, this paper proposes a passive 3D reconstruction method based on binocular stereo matching algorithm. Because the precision of the sparse stereo matching algorithm is not high, and most of them are applied to the indoor scenes, so the region-based dense stereo matching algorithm is adopted, which uses the correlation degree of the gray information between the windows to match, and selects the optimal similarity measure and window size by objective evaluation. After the initial disparity map is obtained, the disparity map is refined by the parallax reliability detection and the sub-pixel interpolation processing. Finally, the principle of similar triangulation is used for 3D reconstruction, and the texture mapping technology is used. The effectiveness of the algorithm is verified by experiments.