利用GSOM网络实现散乱点云的区域分割  

Point-cloud segmentation based on GSOM neural networks

作　　者：余水晶[1] 刘德平[1] 刘晓宇[1] 王莹莹[1] 

机构地区：[1]郑州大学机电一体化研究所,郑州450001

出　　处：《机械设计与制造》2009年第2期219-221,共3页Machinery Design & Manufacture

基　　金：河南省教育厅自然基金项目(200510459060)

摘　　要：提出了一种改进的动态自组织特征映射(GSOM)神经网络算法,实现逆向工程中点云的区域分割,以数据点的坐标、估算出的法矢量和曲率构成的8维向量作为神经网络的输入,在训练中动态生成网络结构,克服了现有SOFM网络需要预先给出分区数目的限制,网络生成的结点数少、聚类速度快,最后通过实例实现了数据点云的区域分割,验证了该方法的正确性。An improved dynamic self-organization feature map of neural networks （GSOM） is put forward. Based on this algorithm,the region division of point clouds in reverse engineering is realized,eight dimensional feature vectors including 3-dimensional coordinate, 3-dimensional normal vector and 2-dimensional curvature are taken as input of the networks ,the framework of networks is constructed dynamically in model training. Compared with other traditional methods ,this novel method has no necessary to specified the number of neural network segmentation in advance,it has less knots and converges more quickly. The validity of this method is proved by the real example.

关 键 词：逆向工程 神经网络 数据分割 

分 类 号：TH16[机械工程—机械制造及自动化] TP391[自动化与计算机技术—计算机应用技术]