摘要
利用径向基函数(RBF)神经网络算法对来自曲面原型的散乱测量数据点进行预拟合,通过径向基函数与隐层权值的线性组合来映射自由曲面的数学模型,并将该数学模型转化为双三次B样条曲面,既精确、快捷地反求了曲面原型,符合光顺、抗噪的要求,又具备CAD/CAM系统中几何模型实用、标准的表达形式,符合可编辑、可交换的要求·
The free-form surface reconstruction is one of the important skills of reverse en gineering. A novel method for B-spline surface reconstruction by artificial neu ral network was introduced. An effective radius-basis-function (RBF) network m odel was applied in pre-fitting of scattered data from an original surface. Usi ng a multi-step B-spline surface approximation algorithm,the mathematical mode l called as a RBFNN model,was transformed to a bicubic B-spline surface. This model can reconstruct original surfaces precisely and quickly, fitting to fairin g demand. The reconstructed surfaces have an available and standard format in CA D/CAM system, fitting to editable and exchangeable demand. Some key techni ques were analyzed in detail. The approach is feasible and valuable for practica l applications.
出处
《东北大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2003年第6期556-559,共4页
Journal of Northeastern University(Natural Science)
基金
国家高技术研究发展计划项目(2001AA412020)
辽宁省科学技术计划项目(2001216001)
关键词
反向工程
散乱数据
曲面重构
径向基函数
神经网络
B样条
reverse engineering
scattered data
surface reconstruction
RBF
neural network , B-spline
