A tetrahedral mesh generation method based on which the mesh density could be adjusted with the shape of the object was developed in the paper. The BCC lattice was used to initialize the mesh space. According to the curvature and the distance from the lattice to the surface of the object

the LEPP algorithm is adopted to refine the initialized mesh from the outer to the inner with different scales. Then all the lattices that stride the surface were adjusted to fit the shape of the object. At last the quality function navigated Laplacian smoothing and Edge Collapse methods were used to optimize the quality of the meshed tetrahedral and the meshed model was obtained. Result shows the method improves the configuration of the mesh while the quality of the tetrahedral is also guaranteed. The method is tested to be useful in the finite element modeling based on 3D computer tomography images or surface model.