基于支持向量机的图像语义提取研究  

Semantic Extraction of Image Using Support Vector Machines

作　　者：谢刚[1] 古赟[1] 王芳[1] 雷少帅[1] 

机构地区：[1]太原理工大学信息工程学院,大原030024

出　　处：《太原理工大学学报》2011年第6期563-565,570,共4页Journal of Taiyuan University of Technology

基　　金：山西省回国留学人员资助基金(2009-31);山西省重大产业技术开发项目([2009]446号)

摘　　要：通过研究自然景观图案的语义分类,分析了不同的核函数和参数优化算法对图像语义分类性能的影响,并用自然景观图片进行了验证。实验结果表明:当核函数为RBF且参数采用网格搜索优化时,SVM的分类效果最优,可实现对自然景观图像的准确分类。此结论对SVM在图像语义分类中的推广应用具有指导意义。SVM has been paid more and more attention for its good classification ability in image semantic classification field.As the classification performance is decided by SVM kernel functions and parameters together,this paper analyzed the influences of different kernel functions and parameter optimization algorithms on image semantic classification performance,and verified the influences with natural landscape pictures.The experimental result shows the SVM method obtained the best semantic classification effect when the RBF kernel function was choosen and parameters were optimized by grid search method.This conclusion has reference meaning in image semantic classification.

关 键 词：支持向量机 语义分类 核函数 网格搜索 粒子群优化 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]