基于八叉树与RBF神经网络的曲面三角网格生成

The Surface Delaunay Triangulation Based on Octree and RBF Neural Network

采用八叉树空间分解法生成曲面离散点集,再由点的最近邻域搜索得到三维散乱点集的Delaunay三角剖分算法,并通过对输入的散乱点集进行数据平滑预处理,确定特征点和提取特征线,从神经网络的权值矩阵直接计算曲线控制顶点或曲面的控制网格,最后结合RBF神经网络,提出了一种快速、有效的曲面三角网格直接生成法,其网格生成质量较好.该方法对服装曲面造型和曲面裁片具有重要的意义.

In this paper,the scatter points on surface are reached by octree special discrete method,of which the Delaunay triangulation algorithm has been obtained through the search of the point＇s nearest region.Then we pre-treat these points to data smoothing,make sure the characteristic points,extract the edge line and directly calculate the curve controllable vertices or meshes of the surface from the neural network weight-value.At last,combined the RBF neural network,a directly fast and efficient triangulation algorithm with excellent mesh quality,which is important to DressingSim system.