计算机工程 ›› 2018, Vol. 44 ›› Issue (11): 222-227.doi: 10.19678/j.issn.1000-3428.0048918

• 图形图像处理 • 上一篇    下一篇

基于改进核相关滤波器的目标跟踪算法

江维创,张俊为,桂江生   

  1. 浙江理工大学 信息学院,杭州 310018
  • 收稿日期:2017-10-11 出版日期:2018-11-15 发布日期:2018-11-15
  • 作者简介:江维创(1992—),男,硕士研究生,主研方向为计算机视觉、图像处理、机器学习;张俊为,硕士研究生;桂江生,副教授、博士。
  • 基金项目:

    国家自然科学基金(61105035);浙江省重大科技专项重点工业项目(2014C01047)。

Target Tracking Algorithm Based on Improved Kernel Correlation Filter

JIANG Weichuang,ZHANG Junwei,GUI Jiangsheng   

  1. School of Informatics and Electronics,Zhejiang Sci-Tech University,Hangzhou 310018,China
  • Received:2017-10-11 Online:2018-11-15 Published:2018-11-15

摘要:

在跟踪场景中,目标自身和背景会发生不可预测的变化,给目标跟踪带来较大困难。针对该问题,建立一种基于核相关滤波器(KCF)处理遮挡的跟踪算法。根据前向分类器响应最大值的分布特征建立遮挡处理模型,采用阈值方法进行遮挡检测,在目标受到遮挡之后通过块区域螺旋搜索方法进行目标搜索,在目标搜索过程中计算滑动框的响应判定是否为目标。在OTB测试序列集的测试结果表明,与Staple、DSST、KCF、SST算法相比,该算法在跟踪准确度上比次优方法提高6.1%,在跟踪成功率上比次优方法提高1.5%。

关键词: 核相关滤波器, 目标跟踪, 分类器响应, 遮挡检测, 目标搜索

Abstract:

In the real tracking scene,the unpredictable changes of the target itself and the background bring great difficulty to the target tracking.This paper proposes a target tracking algorithm based on Kernel Correlation Filter (KCF) for occlusion.The algorithm establishes an occlusion processing model based on the distribution characteristics of the forward classifier response maximum value,and uses the threshold method to perform occlusion detection.After the target is occluded,the target search is performed by the block region helix search method,and the response of the sliding box is calculated in the target search process to determine whether the target is found.The algorithm is tested on the OTB test sequence and compared with the Staple,DSST,KCF,SST algorithms.Results show that the proposed algorithm improves the tracking accuracy by 6.1% compared with the suboptimal algorithm,and increases the tracking success rate by 1.5% compared with the suboptimal algorithm.

Key words: Kernel Correlation Filter (KCF), target tracking, classifier response, occlusion detection, target search

中图分类号: