加权距离判别分析及其在模式识别中的应用

Discriminant Analysis Based Weighted Mahalanobis Distance and Its Application On Pattern Recognition

文中提出了一种新的加权马氏距离判别分析法,并将其运用于基于模式识别的在线检验系统对零件变形量合格与否进行检验。重点介绍了基于"分布均匀度"来计算马氏距离各分量权值的新算法,同时讨论了应用此判别方法建立零件变形量检验模型的有关问题。文中还给出了应用基于判别分析的零件变形量在线检验系统对一种扁平细小零件进行建模的步骤和检验的结果。结果表明:基于判别分析的零件检验方法是可行的和有效的,并已显示出良好的工程应用前景。

This paper presents a new discriminant analysis method based on weighted Mahalanobis distance and applied it to the on-line-inspection system based pattern recognition for testing the deformation of mechanical parts. It mainly introduces a new algorithm based on a measure, called distributional uniformity,to calculate the weights of each index on MD,and discusses its application that is how to create mathematical modelling for inspecting parts. This paper also gives the modeling steps of on-line inspection system for a flat small part based on this discriminant analysis and the test result. The results show that this inspection method for mechanical elements based discriminant analysis is feasible and effctive, and has demonstrated a good prospect of engineering application.