Abstract: An STL model is divided into many sub-domains with features preserved at sub-domain boundaries, and then a parametric plane is constructed for each sub-domain, where a mapping based advancing front mesh generation algorithm is applied to generating high quality triangular mesh. In order to remove mapping distortion between the real and parametric spaces, a G1 continuous triangular Bernstein-Bézier surface is reconstructed for each sub-domain to help calculate the Riemannian metrics of surface points. Mesh sizes of boundary sample points are specified by automatic calculations of curvature and proximity features of surface models. And a size field is built on the Delaunay triangulation of sample points. This field is smoothed to prevent unaccepted large gradation between mesh regions with different element size values. Numerical experiments demonstrate that the proposed algorithm can produce high qualified adaptive mesh for complex STL surfaces.

Key words: Adaptive, Mesh size, Riemannian metric, STL, Surface mesh generation