基于形态谱的磨粒图像特征参数提取方法  

Extracting Feature Parameter of Wear Particles Image Based on Morphological Spectrum

作　　者：王国德[1] 张培林[1] 李兵[1] 柳成成[1] 张安成[2] 

机构地区：[1]军械工程学院一系 [2]77611部队

出　　处：《润滑与密封》2011年第4期30-32,43,共4页Lubrication Engineering

基　　金：清华大学摩擦学国家重点实验室开放基金项目(SKLTKF09B06)

摘　　要：为提高磨粒识别的精度,提出一种基于形态谱磨粒图像特征参数提取新方法,给出磨粒图像的归一化形态谱的计算方法,并将磨粒的形态谱作为其特征向量,采用径向基函数神经网络对磨粒进行自动识别。结果表明:利用磨粒的形态谱实现了对球形磨粒、切削磨粒、严重滑动磨粒、疲劳剥块4种典型磨粒的分类识别,磨粒的形态谱可以作为磨粒的有效特征参数。In order to increase the identification accuracy of wear particles, a new method was proposed to extract the feature parameters of the wear particle image according to morphological spectrum. The method to compute normalized mor- phological spectrum of wear particles image was given, a radius basis function（RBF） neural network was introduced to realize the automatic recognition of wear particles using the normalized morphological spectrum as the input vectors. The re- suits show that spherical, cutting, severe sliding and fatigue spall particles can be well identified according to morphological spectrum, the normalized morphological spectrum can be used as the comprehensive feature parameter in wear particle image analysis.

关 键 词：磨粒图像 特征提取 形态谱 径向基函数神经网络 

分 类 号：TK401.2[动力工程及工程热物理—动力机械及工程]