计算机应用 ›› 2016, Vol. 36 ›› Issue (9): 2597-2600.DOI: 10.11772/j.issn.1001-9081.2016.09.2597

• 虚拟现实与数字媒体 • 上一篇    下一篇

基于特征融合与核局部Fisher判别分析的行人重识别

张耿宁, 王家宝, 李阳, 苗壮, 张亚非, 李航   

  1. 解放军理工大学 指挥信息系统学院, 南京 210007
  • 收稿日期:2016-02-29 修回日期:2016-04-14 出版日期:2016-09-10 发布日期:2016-09-08
  • 通讯作者: 王家宝
  • 作者简介:张耿宁(1991-),男,广东揭东人,硕士研究生,主要研究方向:行人重识别、计算机视觉;王家宝(1985-),男,安徽肥西人,讲师,博士,CCF会员,主要研究方向:机器学习、计算机视觉;李阳(1984-),男,河北廊坊人,讲师,博士研究生,主要研究方向:机器学习、计算机视觉;苗壮(1976-),男,辽宁辽阳人,副教授,博士,主要研究方向:智能信息处理;张亚非(1955-),男,河北易县人,教授,博士,主要研究方向:人工智能;李航(1983-),男,江苏南京人,博士研究生,CCF会员,主要研究方向:计算机视觉。
  • 基金资助:
    国家自然科学基金资助项目(61402519);江苏省自然科学基金资助项目(BK20140071,BK2012512)。

Person re-identification based on feature fusion and kernel local Fisher discriminant analysis

ZHANG Gengning, WANG Jiabao, LI Yang, MIAO Zhuang, ZHANG Yafei, LI Hang   

  1. College of Command Information Systems, PLA University of Science and Technology, Nanjing Jiangsu 210007, China
  • Received:2016-02-29 Revised:2016-04-14 Online:2016-09-10 Published:2016-09-08
  • Supported by:
    This work is partially supported by the National Natural Science Foundation of China (61402519) and the Jiangsu Provincial Nature Science Foundation (BK20140071, BK2012512).

摘要: 行人重识别精度主要取决于特征描述和度量学习两个方面。在特征描述方面,现有特征难以解决行人图像视角变化的问题,因此考虑将颜色标签特征与颜色和纹理特征融合,并通过区域和块划分的方式提取直方图获得图像特征;在度量学习方面,传统的核局部Fisher判别分析度量学习方法对所有查询图像统一映射到相同的特征空间中,忽略了查询图像不同区域的重要性,为此在核局部Fisher判别分析的基础上对特征进行区域分组,采用查询自适应得分融合方法来描述图像不同区域的重要性,由此实现度量学习。在VIPeR和iLIDS数据集上,实验结果表明融合后的特征描述能力明显优于原始特征,同时改进的度量学习方法有效提高了行人重识别精度。

关键词: 行人重识别, 颜色标签, 特征融合, 度量学习, 核局部Fisher判别分析

Abstract: Feature representation and metric learning are fundamental problems in person re-identification. In the feature representation, the existing methods cannot describe the pedestrian well for massive variations in viewpoint. In order to solve this problem, the Color Name (CN) feature was combined with the color and texture features. To extract histograms for image features, the image was divided into zones and blocks. In the metric learning, the traditional kernel Local Fisher Discriminant Analysis (kLFDA) method mapped all query images into the same feature space, which disregards the importance of different regions of the query image. For this reason, the features were grouped by region based on the kLFDA, and the importance of different regions of the image was described by the method of Query-Adaptive Late Fusion (QALF). Experimental results on the VIPeR and iLIDS datasets show that the extracted features are superior to the original feature; meanwhile, the improved method of metric learning can effectively increase the accuracy of person re-identification.

Key words: person re-identification, Color Name (CN), feature fusion, metric learning, kernel Local Fisher Discriminant Analysis (kLFDA)

中图分类号: