There are many of 3D model reconstruction software based on cloud computation technology released recently. Photosynth is the earliest one which based on Scaleinvariant feature transform (SIFT) and Structure from Motion (SfM). SIFT gives function of invariance of scale, rotation, perspective and lighting for feature point extraction. This kind of approach is automatic feature point extraction and matching. SfM can restore camera station and acquire space coordinates by continuous photos. In 2008, Microsoft released a free software named Photosynth which combined with SIFT and SfM technology. It can analyze digital photographs and generate a 3D point cloud of a photographed object.In this paper, the accuracy was estimated by the index of RMSE of the point cloud generated from Photosynth, Lidar, and close-range photogrammetry by using some empirical data. The results indicate that the 3D positioning accuracy of Photosynth approach under well control is about ±0.027m, ±0.065m, and ±0.012m respectively. The result is used to compare to that of close-range Photogrammetry and Lidar. We found out the restriction and the problem in this method either.