基于SVM的粉末冶金零件的多类分类器的研究  

Research on Multi-class SVM Classifier Based on Powder Metallurgy Parts

作　　者：张小洁[1] 张艳[1] 林育阳 

机构地区：[1]陕西工业职业技术学院,陕西咸阳712000 [2]陕西省机械研究院,陕西咸阳712000

出　　处：《计算技术与自动化》2017年第2期33-36,共4页Computing Technology and Automation

基　　金：陕西工业职业技术学院科研基金项目(ZK16-05)

摘　　要：传统的SVM特别适合解决两类分类问题,而对于多类分类,则需将其转化为多个两类分类问题,相应地需要构造多个两类子分类器,这样不但使得分类器结构复杂,而且分类速度受到很大的影响。为了快速地进行多类分类,本文使用LIBSVM中的svmtrain实现对训练数据集的训练,从而获取SVM多分类模型,利用获取的模型进行测试与预测,不仅使得子分类器数目大大减少,而且使分类速度明显提高。最后从粉末冶金零件图库中选取的8张图像进行了分类实验,取得较好的分类结果。The traditional SVM is very suitable for solving two class classification problems, and for multi class classification, will be transformed into multiple two class classification problems, accordingly need to construct a plurality of two types of sub classifiers. This not only makes the classifier complexity and classification speed have great influence. In order to fast multi class classification. In this paper,we use libsvm svmtrain training on the training data set, in order to obtain the SVM multi classification model, are utilized to obtain the model of test and forecast, not only makes the number of sub classifiers is greatly reduced,and the classification rate is improved significantly. Finally, the 8 images selected from the powder metallurgy parts library are classified,and the results are obtained.

关 键 词：SVM 图像分类 粉末冶金零件 多类分类器 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]