作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2018, Vol. 44 ›› Issue (12): 264-270. doi: 10.19678/j.issn.1000-3428.0049136

• 图形图像处理 • 上一篇    下一篇

相关滤波跟踪算法的特征分析

刘霖枫,孔繁锵,严小乐,沈秋   

  1. 南京航空航天大学 航天学院,南京 210016
  • 收稿日期:2017-10-31 出版日期:2018-12-15 发布日期:2018-12-15
  • 作者简介:刘霖枫(1992—),男,硕士研究生,主研方向为视觉目标;孔繁锵,讲师;严小乐,硕士;沈秋,讲师。
  • 基金资助:

    国家自然科学基金(61201365,61401200)。

Feature Analysis in Correlation Filtering Tracker Algorithm

LIU Linfeng,KONG Fanqiang,YAN Xiaole,SHEN Qiu   

  1. College of Astronautics,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China
  • Received:2017-10-31 Online:2018-12-15 Published:2018-12-15

摘要:

特征提取是影响相关滤波跟踪算法性能的关键步骤,但现有特征提取方法没有考虑特征优化和特征融合的影响。为此,利用线性核函数下的跟踪表现可直观反映特征描述能力的特点,采用OTB和VOT 2种评价标准,通过大量实验对常用的灰度、形状、纹理和颜色特征进行系统的对比与评价。同时,分析多种跟踪场景下各特征的适用性和互补性,为跟踪算法中单特征选择或多特征融合提供理论和实验依据。实验结果表明,在复杂场景中,形状、纹理和颜色特征具有较好的互补性。

关键词: 目标跟踪, 相关滤波, 特征选择, 特征适用性, 特征互补性

Abstract:

Feature extraction is one of the most important steps in correlation filtering based on tracker algorithm,but existing feature extraction methods do not take into account the effects of feature optimization and feature fusion.As the tracking performance under linear kernel can directly reflect the characteristics of feature description,two criteria which are OTB and VOT are used to evaluate the description capability and complementarity of grayscale,shape,texture and color features in multiple tracking scenarios.The applicability and complementarity of each feature in various tracking scenarios are analyzed,which provides theoretical and experimental basis for single feature selection or multi-feature fusion in tracking algorithm.Expenmental results conducted show that shape,texture and color features can complement each other in complex scenes.

Key words: object tracking, correlation filtering, feature selection, feature applicability, feature complementarity

中图分类号: