Research On Sheet Metal Surface Fitting And Region Segmentation Of Point Cloud Data Based On The Curvature Characteristics

Sheet metal has complex product structure and is mostly designed and processed by some important structural features. A major part of the metal surface is composed of some relatively simple analytical surfaces such as plane, spherical and cylindrical surface, and these surfaces generally must to be satisfied with certain geometric constraints. Sheet metal surface reconstruction is not only to reproduce the structural features, but also to reproduce constraint relationship between the surfaces. Sheet metal surface reconstruction with satisfied constraints can improve the accuracy of surface reconstruction, and be more suitable for the designer’s intent. So the reconstructed sheet metal surface model can be used for subsequent retrofit design, mold design, production and processing efficiently.This paper focuses on the structural characteristics of the sheet metal, and sheet metal surface reconstruction is divided into four stages:preprocessing of point cloud data, region segmentation of point cloud data, recognition and constraint recognition of quadratic surface, and finally surface fitting based on the characteristics and constraints. The main work is as follows:1. Preprocessing of point cloud data. Estimate the normal vector and curvature of scattered point cloud data, and correct the normal vector before curvature estimating. Estimate of the normal vector and curvature of point data near the boundary is sometimes not accurate, which will affect the surface feature identification. In order to avoid this problem, point cloud data of the transition region is extracted before the regional segmentation.2. Region segmentation of point cloud data. Apply the eight-dimensional feature vector component, which contain the estimated normal vector, Gaussian curvature and mean curvature and coordinates of the pre-treated point cloud data to an algorithm based on Genetic Fuzzy Clustering to achieve the segmentation of point cloud data.3. Quadratic surface feature recognition and constraint recognition. The initial characteristic parameters of a plane are obtained by the least squares method, and use the discriminant method to determine the initial parameters of other quadratic surfaces. For the surface whose feature is not complete or whose curvature is rather small, PSO based on the discriminant method is applied to determine the type of surface and complete surface fitting. The relationship of constraints between the surfaces can be identified by reference feature off-line parameterization method, at the same time specific method of the constraint recognition method is provided.4. Surface fitting based on the characteristics and constraints. Surface optimal fitting based on the characteristics and constraints is constructed. Constrained PSO method is used to solve optimization fitting model. It can reduce the difficulty of solving and increase the accuracy of surface fitting. Finally the sheet metal surface model is established after extension, intersection, cutting, transition and other regular operations of the reconstructive surface.Using CAA, the secondary development tool of CATIA, a system of sheet metal surface reconstruction is programmed. Two sheet metal models are used for surface fitting to prove the practicality of the system.