基于SVM的图像分类  

CLASSIFYING IMAGES WITH SVM METHOD

作　　者：王亮申[1] 欧宗瑛[2] 朱玉才[1] 侯杰[1] 于京诺[1] 

机构地区：[1]烟台师范学院,山东烟台264025 [2]大连理工大学,辽宁大连116023

出　　处：《计算机应用与软件》2005年第5期98-99,126,共3页Computer Applications and Software

摘　　要：SVM能够解决线性及非线性分类问题,以较少的支持向量确定分类面,对样本数量及维数不敏感。基于直方图及惯性比确定图像特征能够保证平移、旋转和尺度不变性,是最优分类面确定的可靠保证。利用SVM进行图像分类,充分体现SVM理论的实用价值。Linear and nonlinear classifications c an be made with SVM,and the hyperplane can be defined by a few support vectors.SVM is insensitive to dimensions and numbers of training set.The features based on histogram and ratio of inertial moment can keep their scale-invariance,rotation -invariance and translation-invariance,and get optimal hyperplane.The classifi cation to images with SVM proves that the SVM is a correct method practically an d experimentally.

关 键 词：图像分类 尺度不变性 最优分类面 SVM理论 分类问题 支持向量 样本数量 图像特征 实用价值 非线性 直方图 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN911.6[自动化与计算机技术—计算机科学与技术]