增量深度学习目标跟踪

程帅; 孙俊喜; 曹永刚; 赵立荣

doi:10.3788/OPE.20152304.1161

您当前的位置：

首页 >

文章列表页 >

增量深度学习目标跟踪

信息科学 | 更新时间：2020-08-13

- 增量深度学习目标跟踪
- Target tracking based on incremental deep learning
- 光学精密工程 2015年23卷第4期页码：1161-1170
- 作者机构：
  
  1. 中国科学院长春光学精密机械与物理研究所,吉林长春,中国,130033
  2. 长春理工大学电子信息工程学院,吉林长春,130022
  3. 东北师范大学计算机科学与信息技术学院,吉林长春,130117
- 作者简介：
- 基金信息：
  
  吉林省科技厅资助项目(No.20090512();20100312)()
- DOI：10.3788/OPE.20152304.1161
  中图分类号： TP391.4
- 收稿日期：2014-10-09，
  
  修回日期：2014-12-03，
  
  纸质出版日期：2015-04-25
- 稿件说明：
移动端阅览
程帅, 孙俊喜, 曹永刚等. 增量深度学习目标跟踪[J]. 光学精密工程, 2015,23(4): 1161-1170

CHENG Shuai, SUN Jun-xi, CAO Yong-gang etc. Target tracking based on incremental deep learning[J]. Editorial Office of Optics and Precision Engineering, 2015,23(4): 1161-1170
程帅, 孙俊喜, 曹永刚等. 增量深度学习目标跟踪[J]. 光学精密工程, 2015,23(4): 1161-1170 DOI： 10.3788/OPE.20152304.1161.

CHENG Shuai, SUN Jun-xi, CAO Yong-gang etc. Target tracking based on incremental deep learning[J]. Editorial Office of Optics and Precision Engineering, 2015,23(4): 1161-1170 DOI： 10.3788/OPE.20152304.1161.

摘要

由于现有目标跟踪算法在复杂环境下易发生目标漂移甚至跟踪丢失

故本文提出了以双重采样粒子滤波为框架

基于增量深度学习的目标跟踪算法。该算法在粒子滤波中引入粒子集规模自适应调整的双重采样来解决粒子衰减及贫化问题

并利用无监督特征学习预训练深度去噪自编码器以克服跟踪中训练样本的不足。将深度去噪自编码器应用到在线跟踪中

使提取的特征集合能够有效表达粒子图像区域。在深度去噪自编码器中添加了增量特征学习方法

得到了更有效的特征集以适应跟踪过程中目标外观变化。该方法还用线性支持向量机对特征集合进行分类

提高对粒子集合的分类精度

以得到更精确的目标位置。在复杂环境下对不同图片序列进行的实验表明:该算法的跟踪综合评价指标为94%、重叠率为74%

平均帧率为13 frame/s。与现有的跟踪算法相比

本算法有效地解决目标漂移甚至跟踪丢失问题

并且对遮挡、相似背景、光照变化、外观变化具有更好的鲁棒性及精确度。

Abstract

As current tracking algorithms lead to target drift or target loss in the complex environment

a tracking algorithm based on the incremental deep learning was proposed under a double-resampling particle filter framework. To solve the problem of particle degradation and depletion

the double-resampling method was introduced to adapt to the particle size in particle filtering and a Stacked Denoising Autoencoder(SDAE) was pre-trained by the unsupervised feature learning to alleviate the lack of training samples in visual tracking. Then

the SDAE was applied to online tracking

so that the extracted feature sets could express the region image representations of the particles effectively. The incremental feature learning was introduced to the encoder of SDAE

the feature sets were optimized by adding new features and merging the similar features to adapt to appearance changes of the moving object. Moreover

a support vector machine was used to classify the features then to improve the classification accuracy of the particles and to obtain a higher tracking precision. According to the results of experiments on variant challenging image sequences in the complex environment

the

-measure and the overlapping ratio of the presented algorithm are 94%

74%

respectively and the average frame rate is 13 frame/s. Compared with the state-of-the-art tracking algorithms

the proposed method solves the problems of target drift and target loss efficiently and has better robust and higher accuracy

especially for the target in the occlusions

background clutter

illumination changes and appearance changes.

关键词

Keywords

references

郭敬明, 何昕, 魏仲慧. 基于在线支持向量机的Mean Shift彩色图像跟踪[J]. 液晶与显示, 2014, 29(1): 120-128. GUO J M, HE X, WEI ZH H. New mean shift tracking for color image based on online support vector machine [J]. Chinese Journal of Liquid Crystals and Displays, 2014, 29(1): 120-128. (in Chinese)

WU Y, LIM J, YANG M H. Online object tracking: A benchmark[C]. 2013 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), Portland, 2013: 2411-2418.

ROSS D A, LIM J, LIN R S, et al. Incremental learning for robust visual tracking [J]. International Journal of Computer Vision, 2008, 77(1-3): 125-141.

李静宇, 王延杰. 基于子空间的目标跟踪算法研究[J]. 液晶与显示, 2014, 29(4): 617-622. LI J Y, WANG Y J. Subspace based target tracking algorithm [J]. Chinese Journal of Liquid Crystals and Displays, 2014, 29(4): 617-622. (in Chinese)

BABENKO B, YANG M H, BELONGIE S. Robust object tracking with online multiple instance learning [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011, 33(8): 1619-1632.

陈东成, 朱明, 高文, 等. 在线加权多示例学习实时目标跟踪[J]. 光学精密工程, 2014, 22(6): 1661-1667. CHEN D CH, ZHU M, GAO W, et al.. Real-time object tracking via online weighted multiple instance learning [J]. Opt. Precision Eng., 2014, 22(6): 1661-1667. (in Chinese)

GRABNER H, GRABNER M, BISCHOF H. Real-time tracking via on-line boosting[C]. Proceedings of British Machine Vision Conference, Edinburgh, Scotland: BMVA, 2006: 47-56.

GRABNER H, LEISTNER C, BISCHOF H. Semi-supervised on-line boosting for robust tracking[C]. Proceedings of European Conference on Computer Vision, Berlin, Germany: Springer, 2008: 234-247.

颜佳, 吴敏渊. 遮挡环境下采用在线Boosting的目标跟踪[J]. 光学精密工程, 2012, 20(2): 439-446. YAN J, WU M Y. On-line boosting based target tracking under occlusion[J]. Opt. Precision Eng., 2012, 20(2): 439-446.(in Chinese)

KALAL Z, MIKOLAJCZYK K, MATAS J. Tracking-learning-detection [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34(7): 1409-1422.

WANG N Y, YEUNG D Y. Learning a deep compact image representation for visual tracking[C]. Proceedings of Twenty Seventh Annual Conference on Neural Information Processing Systems, Lake Tahoe, USA: Nevada, 2013: 5-10.

VINCENT P, LAROCHELLE H, LAJOIE I, et al.. Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion [J]. Journal of Machine Learning Research, 2010, 11: 3371-3408.

BURGES C J C. A tutorial on support vector machines for pattern recognition[J]. Data Mining and Knowledge Discovery, 1998, 2(2): 121-167.

ROB H ALAN F. Discriminatively trained particle filters for complex multi-object tracking[C]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Miami, FL: IEEE, 2009: 240-247.

李天成, 孙树栋. 采用双重采样的移动机器人Monte Carlo定位方法[J]. 自动化学报, 2010, 36(9): 1279-1286. LI T CH, SUN SH D. Double-resampling based Monte Carlo localization for mobile robot[J]. Acta Automatica Sinica, 2010, 36(9): 1279-1286. (in Chinese)

HINTON G E, SALAKHUTDINOV R R. Reducing the dimensionality of data with neural networks [J]. Science, 2006, 313(5786): 504-507.

ZHOU G Y, SOHN K, LEE H. Online incremental feature learning with denoising autoencoders[J]. Journal of Machine Learning Research-Proceedings Track, 2012: 1453-1461.

TORRALBA A, FERGUS R, FREEMAN W T. 80 million tiny images: A large data set for nonparametric object and scene recognition [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2008, 30(11): 1958-1970.

ZHANG K, ZHANG L, YANG M H. Real-time compressive tracking[C]. Proceedings of European Conference on Computer Vision, Florence, Italy, 2012, 3:864-877.

DINH T B, VO N, MEDION G. Context tracker: Exploring supporters and distracters in unconstrained environments[C]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Providence, RI:IEEE, 2011: 1177-1184.

ADAM A, RIVLIN E, SHIMSHONI I. Robust fragments-based tracking using the integral histogram[C]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, New York, USA: IEEE, 2006: 798-805.

COMANICIU D, RAMESH V, MEER P. Kernel based object tracking[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2003, 25(5): 564-577.

浏览量

376

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

数据驱动的多模复合制导信息融合及其试验验证

基于多分支残差注意力网络的水下图像增强

基于策略梯度和伪孪生网络的异源图像匹配

面向点云分类和分割的形状自适应特征聚合网络

基于机载的红外动态目标视频实时超分辨率重建