A new method that can acquire 3D models from real objects with complex surfaces is proposed in this paper. This method first gets the source depth images from a TOF-Camera and get the partial meshes of the object after the image preprocessing such as triangulation, segmentation, denoising and smoothing, then aligned these partial meshes using Iterative Closest Point (ICP) algorithm. During the alignment process delete the overlap data to reduce memory usage. After the align process, in order to get an intact model of the object, we combine the portions together and reconstruct the mesh after remove the overlap points of meshes. Experimental results show that the proposed method can obtain models of complex surface objects in a short time and can highly improve the quality of the original data.