文章摘要
陈志旺,刘旺.特征融合和自校正的多尺度改进KCF目标跟踪算法研究[J].高技术通讯(中文),2022,32(4):337~350
特征融合和自校正的多尺度改进KCF目标跟踪算法研究
Research on improved multi-scale KCF target tracking algorithm based on features fusion and self-correction
  
DOI:
中文关键词: 目标跟踪; 相关滤波; 异常检测; 自校正; 尺度自适应; 特征融合
英文关键词: target tracking, correlation filter, abnormal detection, self-correction, scale adaptation, features fusion
基金项目:
作者单位
陈志旺  
刘旺  
摘要点击次数: 506
全文下载次数: 377
中文摘要:
      传统核相关滤波(KCF)目标跟踪算法使用单一特征,不监控跟踪输出,且跟踪框的大小是固定的,在跟踪对象发生尺度变化、遮挡等异常情况下容易导致跟踪失败。针对这一问题,本文提出一种改进的KCF目标跟踪算法。首先,在快速方向梯度直方图(FHOG)特征的基础上级联颜色特征(CN)训练和更新滤波器;其次,利用峰值旁瓣比(PSR)检查跟踪输出,当检测到跟踪输出异常时,启动自校正机制校正跟踪输出,从而准确地重新跟踪到目标。最后,融入尺度滤波器来适应目标尺度的变化。本文对该算法进行了仿真和实物实验,实验结果验证了该算法的有效性。
英文摘要:
      Traditional kernel correlation filter (KCF) target tracking algorithm uses a single feature, does not monitor the output, and the size of the target box is unchangeable. It can usually cause tracking failures for the target under abnormal conditions such as occluded or scaled. To solve the problems, this paper proposes an improved kernel correlation filter target tracking algorithm. Firstly, the cascade of color name (CN) feature and fast histogram of oriented gradients (FHOG) features are applied to train and update filters. Secondly, the tracking output is monitored by the peak-to-sidelobe ratio (PSR) value. When an abnormal tracking output is detected, a self-correction mechanism is activated to accurately re-track the target. Finally, a scale filter is added to adapt to changes in the target scale. Simulation and physical experimental results verify the effectiveness of the proposed algorithm.
查看全文   查看/发表评论  下载PDF阅读器
关闭

分享按钮