电子学报 ›› 2016, Vol. 44 ›› Issue (4): 821-825.DOI: 10.3969/j.issn.0372-2112.2016.04.010

• 学术论文 • 上一篇    下一篇

基于簇相似的多分类器目标跟踪算法

李康, 何发智, 潘一腾, 孙航   

  1. 武汉大学计算机学院, 湖北武汉 430072
  • 收稿日期:2014-11-15 修回日期:2015-07-31 出版日期:2016-04-25
    • 通讯作者:
    • 何发智
    • 作者简介:
    • 李 康 男,1986年9月出生,安徽亳州人.武汉大学在读博士研究生,主要研究方向:图像处理、目标跟踪. E-mail:likang@whu.edu.cn
    • 基金资助:
    • 国家自然科学基金 (No.61472289); 湖北省自然科学基金 (No.2015CFB254)

Multi-Classifier Object Tracking Based on Cluster Similarity

LI Kang, HE Fa-zhi, PAN Yi-teng, SUN Hang   

  1. School of Computer Science, Wuhan University, Wuhan, Hubei 430072, China
  • Received:2014-11-15 Revised:2015-07-31 Online:2016-04-25 Published:2016-04-25
    • Supported by:
    • National Natural Science Foundation of China (No.61472289); Natural Science Foundation of Hubei Province,  China (No.2015CFB254)

摘要:

由于跟踪过程中目标和背景的变化,传统的单分类器跟踪算法学习到大量的非目标信息而导致跟踪精度降低.针对该问题,本文提出使用树形结构保存历史分类器.在每一帧,根据树中路径距离选择分类器集对测试样本分类.提出了一种基于簇相似性比较的分类算法.通过建立以方差为尺度的特征空间,比较测试样本到簇中心的距离计算相似度,快速计算出目标样本.实验表明本算法能够在复杂条件下实现对目标的鲁棒跟踪.

关键词: 目标跟踪, 多分类器, 在线学习, 类哈尔特征

Abstract:

Due to the changes of target and background during tracking,traditional single classifier tracking algorithms learn a lot of non-target information which result in the decrease of tracking accuracy.In this paper,we propose to use tree structure to save former classifiers as a set.In each frame,a subset of classifiers are chosen according to the path in the tree to classify test samples.We propose a classification algorithm based on cluster similarity comparison.A normalized feature space is established according to the variance of the cluster.The target in a new frame could be got by computing the distance between test samples and the center of the cluster.Experiments show that our algorithm could achieve the goal of robust tracking under complicated conditions.

Key words: object tracking, multiple classifiers, online learning, Haar-like feature

中图分类号: