径向基神经网络重建自由曲面的探讨

An Exploration of Freeform Surface Reconstruction by RBF Neural Network

提出了采用神经网络重建自由曲面的方法 ,建立了用于曲面重建的径向基函数神经网络模型 ,提出并论证了神经网络用于密集散乱点曲面重建的方案 .与常规的重构方法对比 ,分析了其优点和关键技术 ,着重讨论了径向基函数神经网络模型 .仿真实验表明 :采用二层的径向基函数网络 ,对单个曲面片的拟合精度和网络训练速度大大优于 BP网 ,完全满足实用要求 ,具有一定的理论与实用意义 .

A novel method for surface reconstruction by artificial neural network is introduced, and an effective radial basis function(RBF) network model as well as its learning algorithm is presented. A program of surface reconstruction from dense scattered points by neural network is illustrated and some key techniques are analyzed in detail. The simulation experiments demonstrate that the fitting precision for surface patches by a two layer RBF network is much better than a BP network, and its training speed is also much faster. This approach is valuable for practical applications.