尺度自适应的铁路异物侵限PSA-Kcf降维跟踪方法  

Scale Adaptive PSA-Kcf Dimension Reduction Tracking Method for Railway Foreign Body Intrusion

作　　者：屈志坚[1] 周锐霖 孙旭兵 袁慎高 赵亮 QU Zhijian;ZHOU Ruilin;SUN Xubing;YUAN Shengao;ZHAO Liang(School of Eletrical and Automation Engineering,East China Jiaotong University,Nanchang 330013,China;CRRC Times Signal & Communication Co.,Ltd.,Changsha 410199,China)

机构地区：[1]华东交通大学电气工程学院,江西南昌330013 [2]中车时代通信信号有限公司,湖南长沙410199

出　　处：《铁道学报》2019年第5期71-81,共11页Journal of the China Railway Society

基　　金：国家自然科学基金(51867009,51567008);江西省杰出青年人才资助计划(20162BCB23045);江西省自然科学基金(20171BAB206044);江西省重点研发计划(20181BBE58010)

摘　　要：针对铁路异物侵限存在尺度上的外观变化,导致现有目标跟踪算法容易学习到过量背景或局部纹理信息,从而引发跟踪框漂移的问题,提出一种融合尺度估计的核相关滤波目标跟踪算法。利用视觉背景提取器ViBe对铁路沿线侵限异物进行检测,通过密集循环采样和尺度金字塔技术分别提取初始化跟踪框的FHOG特征,用来训练一个核相关位置滤波器和一个PCA降维的尺度滤波器,以实现尺度自适应的铁路侵限异物快速跟踪。实验结果表明:PSA-Kcf算法在跟踪精度上优于无尺度估计环节的生成类算法MeanShift和原生核相关滤波算法Kcf,略高于尺度自适应的SA-Kcf和SAMF算法;在跟踪速度上明显快于MeanShift、SA-Kcf和SAMF算法,能达到与Kcf算法相当的快速跟踪效果。Aiming at the appearance of railway foreign body have change in scale,the existing tracker is easy to learn excessive background or local texture information,and leads to the problem of tracking box drift.A kernel-correlation filtering target tracking algorithm mix with scale filter is proposed.Firstly,the visual background extractor is used to identify the foreign body along the railway.The FHOG features of the initial tracking box are extracted by dense cycle sampling and scale pyramid techniques,which is used to train a kernelized correlation position filter and a scale filter with PCA dimension reduction,to achieve scale-adaptive and fast tracking of foreign body.The experimental result shows that the proposed PSA-Kcf algorithm outperforms the Mean Shift algorithm and the native kernel correlation filter algorithm without scale estimation in tracking accuracy,slightly higher than SA-Kcf and SAMF algorithm;the tracking speed is faster than Mean Shift,SA-Kcf and SAMF algorithm.Equivalent to the Kcf algorithm without scale estimation.

关 键 词：异物侵限 视觉背景提取器 密集循环采样 FHOG特征 核相关滤波 PCA降维 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论] U215.8[自动化与计算机技术—计算机科学与技术]