[1]李 敏,吴 莎.基于深度学习的粒子滤波视频目标跟踪算法[J].计算机技术与发展,2020,30(06):23-28.[doi:10. 3969 / j. issn. 1673-629X. 2020. 06. 005]
　LI Min,WU Sha.Particle Filter Video Target Tracking Algorithm Based on Deep Learning[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2020,30(06):23-28.[doi:10. 3969 / j. issn. 1673-629X. 2020. 06. 005]
点击复制

基于深度学习的粒子滤波视频目标跟踪算法()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
30
期数:
2020年06期
页码:
23-28
栏目:
智能、算法、系统工程
出版日期:
2020-06-10

文章信息/Info

Title:
Particle Filter Video Target Tracking Algorithm Based on Deep Learning
文章编号:
1673-629X(2020)06-0023-06
作者:
李 敏; 吴 莎
长安大学 信息工程学院,陕西 西安 710000
Author(s):
LI Min; WU Sha
School of Information Engineering,Chang’an University,Xi’an 710000,China
关键词:
目标跟踪; 粒子滤波; 卷积神经网络; 深度特征; 手工特征
Keywords:
target tracking; particle filter; CNN; depth feature; manual feature
分类号:
TP391. 41
DOI:
10. 3969 / j. issn. 1673-629X. 2020. 06. 005
摘要:
在视频目标跟踪中,由于环境以及目标形变等因素的影响,会导致跟踪算法出现鲁棒性差的问题,针对该问题提出了一种基于预训练卷积神经网络,在粒子滤波框架下将深度特征和手工特征相结合的视觉目标跟踪算法。 该算法通过卷积神经网络对数万张通用目标图像进行离线预训练,得到可以对通用目标进行表示的从简单到复杂的结构性特征,再在粒子滤波跟踪框架下将深度特征和手工特征相结合用于目标跟踪。 同时,该算法以一种懒惰的方式更新跟踪模型,避免了模板频繁更新而导致的昂贵计算代价。 实验结果表明,与已有的传统粒子滤波跟踪方法相比,该方法在现有的跟踪基准测试中显示出优越的性能,在复杂背景、光照以及目标形变等恶劣条件影响下仍能稳定地跟踪目标,具有更强的鲁棒性。
Abstract:
In video target tracking,the tracking algorithm has poor robustness due to the influence of changes in the environment and its objects and other factors. A target tracking algorithm based on pre-trained convolutional neural network and combining depth features with manual features under the framework of particle filter is proposed to solve this problem. In this algorithm, tens of thousands of generic targets images are pre-trained offline by convolutional neural network,and structural features ranging from simple to complex that can represent generic targets can be obtained. Then deep representations and hand-crafted features are combined for target tracking under the particle filter framework. At the same time, the algorithm updates the tracking model in a lazy way and avoids the expensive calculation cost caused by frequent update of templates. The experiment shows that compared with the traditional particle filter trackingmethods,this method has superior performance in the existing tracking benchmark test, and it can still track the target stably under the influence of complex background,illumination,target deformation and other harsh conditions,with stronger robustness.

相似文献/References:

[1]卢笑 孟正大.陪护机器人粒子滤波定位法中重采样算法研究[J].计算机技术与发展,2010,(04):54.
　LU Xiao,MENG Zheng-da.Research of Particle Filter Resampling Algorithm in Indoor Service Robot Localization[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2010,(06):54.
[2]刘翔 吴谨 祝愿博 康晓晶.基于视频序列的目标检测与跟踪技术研究[J].计算机技术与发展,2009,(11):179.
　LIU Xiang,WU Jin,ZHU Yuan-bo,et al.A Study of Object Detecting and Tracking Based on Video Sequences[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2009,(06):179.
[3]雷云 王夏黎 孙华.基于视频的交通目标跟踪方法研究[J].计算机技术与发展,2010,(07):44.
　LEI Yun,WANG Xia-li,SUN Hua.The Research about Transport Target Tracking Based on Video[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2010,(06):44.
[4]谢之宇 蒋晓瑜 汪熙 裴闯.基于多线索融合的目标跟踪算法研究[J].计算机技术与发展,2011,(03):125.
　XIE Zhi-yu,JIANG Xiao-yu,WANG Xi,et al.A Target Tracking Algorithm Research Based on Multi-Cue Fusion[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2011,(06):125.
[5]尤天来 周海徽[].红外目标跟踪技术研究[J].计算机技术与发展,2011,(10):140.
　YOU Tian-lai,ZHOU Hai-hui.Research of Infrared Target Tracking Technology[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2011,(06):140.
[6]龙凤 薛冬林 陈桂明 杨庆.基于粒子滤波与线性自回归的故障预测算法[J].计算机技术与发展,2011,(11):133.
　LONG Feng,XUE Dong-lin,CHEN Gui-ming,et al.Fault Prediction Algorithm Based on Particle Filter and Linear Autoregressive Models[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2011,(06):133.
[7]赵侃 漆德宁.基于UKF滤波的FDOA和TDOA联合定位跟踪算法[J].计算机技术与发展,2012,(05):127.
　ZHAO Kan,QI De-ning.A Tracking TDOA/FDOA Joint Location Algorithm Based on UKF[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2012,(06):127.
[8]姚放吾 许辰铭.基于目标质心的Meanshift跟踪算法[J].计算机技术与发展,2012,(06):104.
　YAO Fang-wu,XU Chen-ming.A Meanshift Tracking Algorithm Based on Centroid[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2012,(06):104.
[9]陈延利 施永豪.运动目标检测与跟踪的DSP实现[J].计算机技术与发展,2012,(08):82.
　CHEN Yan-li,SHI Yong-hao.DSP Realization of Detection and Tracking for Moving Objects[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2012,(06):82.
[10]张璐,张国良,张维平,等.改进IMM算法在机器人目标跟踪中的应用[J].计算机技术与发展,2013,(02):149.
　ZHANG Lu,ZHANG Guo-liang,ZHANG Wei-ping,et al.Application of Improved IMM Algorithm in Robot Target Tracking[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2013,(06):149.
[11]雷飞,孟晓琼,吕露,等. 基于改进的均值漂移算法的运动汽车跟踪[J].计算机技术与发展,2017,27(02):106.
　LEI Fei,MENG Xiao-qiong,LYU Lu,et al. Moving Vehicle Tracking Based on Improved Mean Shift[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2017,27(06):106.
[12]刘少林,朱文球,孙文静,等.基于联合直方图的自适应粒子滤波跟踪算法[J].计算机技术与发展,2018,28(06):106.[doi:10.3969/ j. issn.1673-629X.2018.06.024]
　LIU Shao-lin,ZHU Wen-qiu,SUN Wen-jing,et al.An Adaptive Particle Filtering and Tracking Algorithm Based on Joint Histogram[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2018,28(06):106.[doi:10.3969/ j. issn.1673-629X.2018.06.024]
[13]闾斯瑶,周武能,李龙龙.基于 SAGBA 优化粒子滤波的目标跟踪[J].计算机技术与发展,2020,30(03):36.[doi:10. 3969 / j. issn. 1673-629X. 2020. 03. 007]
　LYU Si-yao,ZHOU Wu-neng,LI Long-long.Target Tracking Based on SAGBA Optimized Particle Filter[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2020,30(06):36.[doi:10. 3969 / j. issn. 1673-629X. 2020. 03. 007]
[14]马媛媛,杨小军.基于多层深度特征的尺度相关粒子滤波算法[J].计算机技术与发展,2021,31(06):169.[doi:10. 3969 / j. issn. 1673-629X. 2021. 06. 030]
　MA Yuan-yuan,YANG Xiao-jun.A Scale Dependent Particle Filter Algorithm Based onMulti-layer Depth Characteristics[J].COMPUTER TECHNOLOGY AND DEVELOPMENT,2021,31(06):169.[doi:10. 3969 / j. issn. 1673-629X. 2021. 06. 030]

更新日期/Last Update: 2020-06-10