摘要
Web服装图像检索是目前的一个热点研究领域,服装图像的自动标注是其中的一项重要研究内容.只有准确地自动标注服装图像,才能实现服装图像的智能化搜索.本文在大量实验的基础上,提出了一种基于多核SVM服装图像自动标注方法,通过提炼服装图像的本质特征并结合交叉验证法调优参数,明显提高了服装图像自动标注的准确率.
At present,clothing web image retrieval is a hot research fields,in which automatic image annotation of clothing is one of the most important contents.Intelligent search on clothing image can be achieved only by labeling clothing image accurately and automatically.In this paper,a clothing automatic image annotation method based on multiple-kernel SVM( SVM) was proposed based on a large number of experiments.The accuracy of clothing automatic image annotation can be improved obviously by refining the essential characteristics of clothing image and tuning parameters using cross validation method.
出处
《北京服装学院学报(自然科学版)》
CAS
北大核心
2017年第2期54-62,共9页
Journal of Beijing Institute of Fashion Technology:Natural Science Edition
基金
教育部人文社科项目(No:12YJA760014)
北京服装学院2016年研究生创新项目(X2016-096)
关键词
服装图像
图像自动标注
SVM
核函数
clothing image
automatic image annotation
SVM
kernel function