Image, video, audio and graphics are information media in multimedia. In order to use and manage them effectively, the contents implied by them are needed to analyze. Currently, content-based image (video) and audio retrieval make some progress. However, there is no a very efficient method to perform a similar graphics retrieval, especially a similar 3D graphics retrieval. In this paper, an algorithm is presented to implement the similar 3D object recognition and retrieval. In this algorithm, 3D features are first obtained after the meshes of a 3D object are reduced through level of detail. Since the dimension of the extracted 3D features is very huge, minimum spanning tree (MST) are used to reduce the features, then the recognition of similar 3D objects are realized by Support Vector Machine (SVM). The proposed algorithm works well when it is used to recognize and retrieve a 3D terrain.