A multi-case Chinese 3D face database Northwestern Polytechnical University 3D (NPU3D) has been built for automatic face recognition experiments and other possible face model applications. NPU3D contains 10500 3D facial surface data corresponding to 300 individuals, and there are 35 different scans per person. Currently, NPU3D database contain the largest cases in multi-scale systematic variations over the pose, facial expression, accessory and occlusion of each person in the world. In this paper, the database description, acquisition schema and the pre-processing methods are provided to help using the data and future extension. As an application, a multi-pose 3D face recognition algorithm is proposed. Experimental results show that the proposed algorithm has a good performance.