基于层次语义的Web服装图像智能采集方法

以大规模智能采集互联网中的服装图像为目的,研究如何利用互联网上服装图像的伴随文本与服装图像概念之间的关联,实现自动采集各语义对应的服装图像。在HITS(Hyperlink-Induced Topic Search)算法的基础上提出一个基于层次语义的图像采集算法SICR(Semantic-based Image Collection Robot)。该算法在层次语义库的支持下,扩充根集与去除链接工厂页面同步进行。在爬取链接网页前,进行锚文字的相似度计算和页面内容的概念分析,舍弃不符合语义的页面,只下载满足语义的服装图像。算法克服了基于文本分析或链接分析的图像自动提取算法的不足,具有较高的准确率和召回率,实验结果证明了SICR算法的有效性。

Hierarchical Semantic-based Web Intelligent Fashion Image Retrieval Method

Aiming at the large-scale automatic collection of fashion images from Web,this paper studied how to use association between the accompany text and the concept of fashion images on Web pages to collect images automatically.Based on the acquisition of semantic-based Web content and the drawbacks of HITS (Hyperlink-induced Topic Search) method,a novel SICR (Semantic-based Image Collection Robot) method was proposed to collect the fashion images from Web.The proposed method (SICR),under the support of the hierarchical semantic library,removes Link Farm page in the expansion of root set,does the similarity calculation for anchor text when crawling link pages.In addition,it makes a brief conceptual analysis of the page content before downloading images.The experimental results on the large-scale dataset have demonstrated that the proposed method can overcome the deficiency of only text or link-based analysis and improve the precision rate and recall rate of fashion image retrieval,experimental results demonstrate the effectiveness of SICR.