Filtering and reduction for 3-dimensional surface modeling of laser line scanning point cloud

With 3D laser scanning technology, it is possible to record clear and abundant surface information of the measuring object, but it also contains a large amount of redundant information. Because of the complexity of measurement environment, the 3D data obtained by camera contains a large amount of noise, which increases the difficulties of 3D visualization, feature extraction and recognition. In this paper, classical 2D filtering algorithm and 3D spatial clustering are combined for applying to 3D point cloud, which can preserve as much detail as possible on the surface of measured object. Then, Non-Uniform Rational B-Splines (NURBS) surfaces are used for reconstructing the surface of the object from filtered point cloud. In order to reduce the computing time in the reconstruction process while reduce the losses of surface information of the object, a simplification algorithm for point cloud that can preserve the geometric features of the object surface is proposed. The proposed algorithm has explicit significance in surface reconstruction of point cloud with noise, feature extraction and recognition in the future work.