Journal of Frontiers of Computer Science and Technology ›› 2020, Vol. 14 ›› Issue (2): 294-306.DOI: 10.3778/j.issn.1673-9418.1810019

Previous Articles     Next Articles

Correlation Filter Tracking Algorithm Using TSK Fuzzy Logic System

CHEN Chen, GAO Yanli, DENG Zhaohong, WANG Shitong   

  1. 1. School of Digital Media, Jiangnan University, Wuxi, Jiangsu 214122, China
    2. Jiangnan Institute of Computing Technology, Wuxi, Jiangsu 214083, China
  • Online:2020-02-01 Published:2020-02-16

TSK模糊逻辑系统相关滤波器跟踪算法

陈晨,高艳丽,邓赵红,王士同   

  1. 1. 江南大学 数字媒体学院,江苏 无锡 214122
    2. 江南计算技术研究所,江苏 无锡 214083

Abstract:

For the problem how to accurately and quickly distinguish the target from the background in the current target tracking field, the core task of most trackers is how to train a discriminant classifier to distinguish between the target and the surrounding environment. At present, the more advanced kernel correlation filter algorithm (KCF) and improved discriminant correlation filter (DCF) can combine the discriminant classifier with the Fourier transform to improve the tracking speed. Some KCF-based optimization algorithms provide solutions to partial tracking problems, such as the KCF algorithm for scale problems and the KCF algorithm for target disappearance. However, existing algorithms still have some room for improvement in improving accuracy. Aiming at this deficiency, a new fuzzy kernel correlation filter (FKCF) is derived using Takagi-Sugeno-Kang fuzzy logic system (TSK-FLS) based on kernel correlation filter. FKCF inherits the characteristics of high speed and small computational complexity of KFC, and further to improve the robustness, replacing the previous simple Gaussian mapping with the fuzzy membership function and introducing the consequent parameters of TSK-FLS in the process of kernel calculation. Thus, FKCF achieves better tracking accuracy than traditional KCF. Extensive experiments are carried out on 50 randomly selected videos on 4 databases such as OTB50. The experimental results show that accuracies of the FKCF on 10 types of common attributes are all improved compared with the traditional KCF.

Key words: tracker, discriminant classifier, Fourier transform, correlation filter, TSK fuzzy logic system

摘要:

针对当前目标跟踪领域中如何准确迅速地区分目标和背景的问题,大部分跟踪器的核心内容是如何训练一个判别分类器来区分目标和周围环境。目前较为先进的核相关滤波器算法(KCF)及其改进后的判别式相关滤波器(DCF)将判别分类器与傅里叶变换相结合来提升跟踪速度,一些基于KCF的优化算法对部分跟踪难题,如针对尺度问题的KCF算法和针对目标消失的KCF算法提出了解决方案。但当前已有算法在提高精度方面仍有一定的提升空间,针对此,在核相关滤波器的基础上,从TSK模糊逻辑系统(TSK-FLS)的角度出发推导出了一种新的模糊核相关滤波器(FKCF)。FKCF继承了前者的高速和计算量小的特性,为了提高鲁棒性,将之前简单的高斯映射换成了模糊隶属度函数,并且在核计算的过程中引入了后件参数。由于这两项改进,使得在跟踪精度方面比KCF更好。将FKCF算法与KCF等相关算法在OTB50等4个数据集中的50个随机选取的视频上进行了实验,10项常见属性上的精度均有提升。

关键词: 跟踪器, 判别分类器, 傅里叶变换, 相关滤波器, TSK模糊逻辑系统