3D models have got widespread attention in graphics currently. Most 3D models are used for visual calculation, but they only contain geometric and appearance properties and the description of high-level semantic features suitable for automatically matching are very little. So, how to reasonably describe 3D models (feature extraction) is a difficult problem of the prime importance. Voxelization just can solve this problem. The voxel model is helpful for regularization of 3D models and is convenient to extract models’ characteristic signals. This paper proposed a voxelization method based on edclidean distance measurement to voxelize models’ vertices, edges and faces. Experiment results indicated that the method is efficient and effective for generating correct entity voxel models.