基于改进的SVM的甲状腺图像检索  

A Thyroid Image Retrieval Algorithm Based on Improved SVMs

作　　者：任小康[1] 白勇峰[1] 范丽[1] 李颜瑞[1] 

机构地区：[1]西北师范大学数学与信息学院,甘肃兰州730070

出　　处：《计算机工程与科学》2011年第1期127-131,共5页Computer Engineering & Science

摘　　要：针对SVM处理大数据量和区分训练集样本属性的重要性差的问题,我们将SVM和粗糙集结合,构造了基于粗糙集与SVM的图像检索相关反馈算法,将其应用于甲状腺CT图像检索。实验结果表明,改进的SVM分类精度可达到92.53%,相比SVM的分类精度(76.58%)提高了15.95%,进而使检索的查准率和查全率也分别提高到89.53%和29.67%。In allusion to SVM,s defects of handling large amount of data and distinguishing the importance of the training set,this paper joins the SVM classifier with the rough sets theory, and constructs an improved image feedback retrieval algorithm based on rough sets and SVMs,which are used to retrieve thyroid CT images .The results show that the improved SVM classifier can get 92.53% accuracy which is about 15.95% higher than 76.58% using SVM,and the retrieval of poor accuracy and recallprecision are also increasedr by 89.53% and 29.67%.

关 键 词：图像检索 粗糙集 支持向量机 相关反馈 甲状腺CT图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]