应用遗传模糊聚类实现点云数据区域分割

为了准确地实现点云数据的区域分割,将基于遗传算法的模糊聚类算法应用于逆向工程中的点云数据区域分割中。首先估算出法矢量、高斯曲率和平均曲率,并与坐标一起组成八维特征向量,用加权距离代替欧氏距离,然后通过遗传算法获得全局最优解的近似解;最后将近似解作为模糊聚类的初始解进行迭代,实现点云数据的区域分割,从而避免传统FCM算法的局部性和对初始解的敏感性,减少了迭代次数。以汽车钣金件为例,证明了应用遗传模糊聚类实现点云数据区域分割的有效性,并验证了该方法能快速、准确地实现点云数据的区域分割。

Genetic fuzzy clustering algorithm for point cloud data segmentation

In order to realize point cloud data segmentation accurately, this paper applied genetic fuzzy clustering algorithm to the point cloud data segmentation in reverse engineering. First, it estimated the normal vector, Gaussian curvature and mean curvature, together with the coordinates of the eight-dimensional feature vector component, using weight distance replaced the Euclidean distance. Through the genetic algorithm, it obtained the approximate solution of the global optimal solution. Finally it used the approximate solution as the initial solutions of fuzzy clustering iteration achieved the point cloud data region segmentation, therefore, avoided the locality and sensitiveness of the initial condition of fuzzy clustering algorithm, at the same time, it reduced the number of iterations. Taking car sheet metal for an example proves the validation of genetic fuzzy clustering algorithm applied to the point cloud data segmentation. And point cloud data can be segmented fast and accurately by this algorithm.