Abstract: A method for reconstructing the inner surface with feature data points acquired by structured light scanning was proposed. Feature data points acquired by structured light scanning are accurate and the distribution density of those data points is nonuniform. Some redundancy points on the measuring direction were removed by sub-sampling and the new points on the scanning direction were added by interpolating. Then the primitive surface of the measured object was reconstructed using the local tangent plane methods with those of preprocessed data points. One local tangent plane for the neighborhood point set of a certain point was constructed and the primitive surface could be approximated by the tangent plane cluster. Based on extracting the isosurface through Marching cubes algorithm, the initial meshes were obtained and the meshes were optimized through an energy function. Then the smooth surface was obtained through Loop subdivision method. It shows that this method can be used for effectively implementing the reconstruction of the inner surface with the nonuniform data points acquiring by structured light scanning.