This paper provides a preprocessing flow for color three-dimension (3D) point cloud according to the characteristics of laser scanning data. The preprocessing methods and their functions are introduced in detail. Automatic system decision and manual polygon selection methods are applied to eliminate unwanted and noise data successfully, which possibly make improper color models reconstructed. A data reduction method is presented based on Grid reduction method considering color-boundary preservation. It can effectively avoid shape and color distortion in model reconstruction. Several experimental results are presented to indicate the feasibility and reliability of these methods, which can be broadly applied in various fields of products design, antiques and art wares digitization, etc.