基于支持向量机的铁谱磨粒自动识别  

Ferrograghy Abrasive Automatic Recognition Based on Support Vector Machine

作　　者：邱丽娟[1,2] 宣征南[1,3] 张兴芳[2] 何照荣[1] 孙志伟[1] 

机构地区：[1]广东石油化工学院机电工程学院,广东茂名525000 [2]太原理工大学化学化工学院,山西太原030024 [3]广东石油化工学院广东省石化装备故障诊断重点实验室,广东茂名525000

出　　处：《广东石油化工学院学报》2015年第3期30-34,共5页Journal of Guangdong University of Petrochemical Technology

基　　金：茂名市科技计划项目(201327);广东省石化装备故障诊断重点实验室开放基金项目(GDUPTKLAB201314)

摘　　要：为了实现支持向量机对磨损磨粒的自动识别,首先对彩色磨粒图像进行图像处理,运用K-均值聚类、区域生长法和数学形态学将磨粒成功地从图像中分割出来;其次根据各类磨粒形貌特征确定磨粒的形状尺寸、纹理特征和颜色特征参数,并采用相应的方法提取这三方面的参数;由于支持向量机参数的选取对其分类效果影响很大,所以采用遗传算法对其参数做进一步的优化;最后以提取的特征参数为基础,运用优化的支持向量机成功地识别了严重滑动磨粒、球形磨粒、切削磨粒、疲劳磨粒和红色氧化物五类特征磨粒,其识别准确率达90%。结果表明该方法切实可行。In order for the support vector machine to automatically recognize abrasive ,firstly ,color abrasive images are processed ,success-fully segmenting abrasive from the image by using K-means clustering ,region growing method and mathematical morphology .Secondly , the shape and size ,texture and color characteristic parameters are determined according to morphologic characteristics of abrasive ,and ap-propriate methods are used to extract the parameters of these three areas .The different support vector machine parameters have a great im-pact on its classification ,so genetic algorithm is used to optimize the parameters further .Finally ,the optimized SVM is used to identify sev-er sliding ,spherical ,cutting ,fatigue and red oxide abrasive ,and the recognition accuracy rate achieves 90% .The results show that this method is feasible .

关 键 词：支持向量机 自动识别 K-均值聚类 遗传算法 

分 类 号：TH117[机械工程—机械设计及理论]