基于SURF算法的服装视觉图像检索研究  

Visual Clothing Image Retrieval Research Based on the SURF Algorithm

作　　者：李克磊[1] 刘正东[1] 

机构地区：[1]北京服装学院计算机信息中心,北京100029

出　　处：《北京服装学院学报（自然科学版）》2014年第3期57-62,共6页Journal of Beijing Institute of Fashion Technology：Natural Science Edition

基　　金：国家科技支撑计划项目(2013BAH41F00;2013BAH41F01);数字与交互媒体北京市重点实验室开放课题;北京服装学院2014年度研究生科研创新项目

摘　　要：提出了基于SURF算法的服装视觉图像检索系统,使用实时、丰富的移动视觉图像进行服装信息的检索.首先对移动端视觉图像进行预处理,然后利用SURF算法完成服装视觉图像特征检测、特征信息描述和特征匹配,从而完成检索.对SURF-36,SURF-64和SURF-100描述向量的匹配效率、匹配准确度等方面进行试验.结果表明,SURF-64能够在效率与准确度之间达到很好的平衡,较快、较好地完成服装视觉图像的检索.Currently,clothing retrieval and recommendation system under the background of mobile internet only allows users to search clothes by using text messages,such as the brand,color and price. This paper tries to use real-time information like characteristics of mobile terminal images to search wanted clothes. Thus,we come up with clothing recommendation system based on mobile image vision by using edge detection retrieval and SURF technology. This system firstly pre-processes clothing images and then uses edge detection technology to locate position of clothing and edge. Finally,it generates clothing image characteristics and completes clothing retrieval matching with SURF algorithm. We test SURF-36,SURF-64 and SURF-100 in such aspects as the matching efficiency and accuracy. The experiments show that SURF-64 can complete clothing image retrieval and matching efficiently and has high matching rate within a certain range of garment samples.

关 键 词：服装 图像检索 SURF 移动视觉 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]