基于Kohonen神经网络的B样条曲面重构.pdf

第 25卷第 9期 计算机应用 V ol. 25 No. 9 2005年 9月 ComputerApplications Sept. 2005 : 1001- 9081( 2005) 09- 2018- 04 Kohonen B 范彦革, 刘旭敏, 陈 婧 (首都师范大学信息工程学院, 北京 100037) ( fanyange@ 126. com ) : 探讨了三维散乱数据点的 由曲面 组织重构方法建立了基于 组织特征映射神经 网络的矩形网格重构模型及其训练算法所建模型利用神经元对曲面散乱数据点的学习和训练来模 拟曲面上点与点之间的内在关系, 节点连接权向量集作为对散乱数据点集的工程近似化并重构曲面 样本点的内在拓扑关系通过该方法不仅能够对无规则散乱数据点进行逼近, 并且通过该方法得到 的曲面也可以作为后继曲面重构的初始曲面仿真实验表明, 所建神经网络模型可实现三维密集无 规则数据点的曲面 组织重构集 压缩于一体 : 曲面重构; B样条曲面; Kohonen神经网络; 组织特征映射 : P183 : A Bsp line surface recon struction based on Kohonen neural network FAN Yange, LIU Xum in, CHEN Jing (College of Inf orm ation Engineering, Cap ital N ormal University, B eij ing 100037, China ) A bstract: he approach to the freeform surface selforgan izing reconstruction for the dense 3D scattered data w as discussed. Based on the selforganizing featurem ap neu ral network, a rectanglem esh reconstruction approach and the train ing algorithm w ere developed. he inherent topologic relations between the scattered points on the surfacew ere learned by the self organ izing feature m ap neural network. he w eight vectors of the neurons on the output layer of the neural newt ork w ere used to approx mi ate the scattered data points. By this approach, not only to approx mi ate the scattered data points and the surface which is reconstructed by thismethod can be as base surface for further process, but also the expermi ent ind icates that by this approach, the reconstruction of the surface and the reduce of the dense scattered data points are comb ined into the same process. he computer smi ulation result show s that th ismethod is effective. K ey w ords: surface reco