特征融合的尺度自适应核相关跟踪算法  

Kernelized correlation tracking algorithm with scale adapting and feature fusion

作　　者：叶秋笛 袁亮[1] 吕凯 YE Qiu-di;YUAN Liang;LYU Kai(School of Mechanical Engineering,Xinjiang University,Urumqi 830047,China)

机构地区：[1]新疆大学机械工程学院,新疆乌鲁木齐830047

出　　处：《计算机工程与设计》2022年第2期420-426,共7页Computer Engineering and Design

基　　金：新疆维吾尔自治区自然科学基金项目(2019D01C02)。

摘　　要：针对KCF算法在跟踪过程中面对复杂场景、尺度变化等问题效果欠佳的问题,提出一个采用KCF跟踪器的特征融合的尺度自适应核相关跟踪算法。提出一种方向梯度直方图和颜色直方图特征的特征融合方法;采用一种含有7个固定尺度因子的尺度池,采用响应最大的尺度更新当下尺度;使用平均相关峰值能量作为模板更新的指标,控制响应更新。将改进算法在OTB-100数据集中与近年来流行跟踪算法进行对比实验,实验结果表明,改进算法在性能上较KCF算法在精度和成功率上分别提高了10.3%和16.3%,可实现目标的鲁棒性跟踪。In view of the poor performance of KCF algorithm in the process of tracking complex scenes and scale variation,a kernel correlation tracking algorithm with scale adapting and feature fusion based on KCF tracker was proposed.A feature fusion method of histogram of oriented gradient and color histogram was proposed.A scale pool containing 7 fixed scale factors was used to update the current scale with the maximum response scale.The average peak-to-correlation energy was used as a template updates index to control response updates.The improved algorithm was compared with the popular tracking algorithms in recent years in the OTB-100 data set.Experimental results show that the improved algorithm improves the accuracy and success rate by 10.3%and 16.3%respectively compared with the KCF algorithm in terms of performance,and the target robust tracking can be achieved.

关 键 词：目标跟踪 相关滤波器 颜色直方图 尺度池 平均相关峰值能量 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]