Aiming at the statistical characteristic function and conventional spherical harmonic functions could not effectively describe 3D rock models, rock spherical harmonic functions were proposed. This algorithm was an improvement on the conventional spherical harmonic functions used in 3D model retrieval. Firstly, the 3D rock model was decomposed into a series of concentric spheres. On each sphere, a spherical function was obtained. By restricting the spheres to different radii, a collection of spherical functions could be obtained. Then, spherical harmonic transform was used on these functions to get morphological characteristics descriptor. Finally, a similarity comparison function was used to measure the similarity between each 3D rock models .Experiments on multi-group sandstone core samples show that the proposed algorithm is efficient in describing 3D rock models, and that the depicting ability of the new algorithm is exacter than the statistical characteristic function and the conventional spherical harmonic functions.
