基于深度稀疏学习的鲁棒视觉跟踪

王鑫; 侯志强; 余旺盛; 戴铂; 金泽芬芬

doi:10.13700/j.bh.1001-5965.2016.0788

基于深度稀疏学习的鲁棒视觉跟踪

doi: 10.13700/j.bh.1001-5965.2016.0788

空军工程大学信息与导航学院, 西安 710077

基金项目:

国家自然科学基金 61473309

国家自然科学基金 61703423

陕西省自然科学基础研究计划 2015JM6269

陕西省自然科学基础研究计划 2016JM6050

详细信息

作者简介:
王鑫男，硕士研究生。主要研究方向：计算机视觉、机器学习

侯志强男，博士，教授，博士生导师。主要研究方向：图像处理、计算机视觉和信息融合

通讯作者:
侯志强, E-mail: hou-zhq@sohu.com

中图分类号: TP391
计量
- 文章访问数: 758
- HTML全文浏览量: 157
- PDF下载量: 572
- 被引次数: 0
出版历程
- 收稿日期: 2016-10-11
- 录用日期: 2017-01-06
- 网络出版日期: 2017-12-20

Robust visual tracking based on deep sparse learning

Information and Navigation College, Air Force Engineering University, Xi'an 710077, China

Funds:

National Natural Science Foundation of China 61473309

National Natural Science Foundation of China 61703423

Natural Science Basic Research Plan in Shaanxi Province 2015JM6269

Natural Science Basic Research Plan in Shaanxi Province 2016JM6050

More Information

Corresponding author: HOU Zhiqiang, E-mail: hou-zhq@sohu.com

摘要

摘要:
视觉跟踪中，高效鲁棒的特征表达是复杂环境下影响跟踪性能的重要因素。提出一种深度稀疏神经网络模型，在提取更加本质抽象特征的同时，避免了复杂费时的模型预训练过程。对单一正样本进行数据扩充，解决了在线跟踪时正负样本不平衡的问题，提高了模型稳定性。利用密集采样搜索算法，生成局部置信图，克服了采样粒子漂移现象。为进一步提高模型的鲁棒性，还分别提出了相应的模型参数更新和搜索区域更新策略。大量实验结果表明：与当前主流跟踪算法相比，该算法对于复杂环境下的跟踪问题具有良好的鲁棒性，有效地抑制了跟踪漂移，且具有较快的跟踪速率。
- 视觉跟踪 /
- 深度学习 /
- 深度稀疏神经网络 /
- 稀疏自编码器 /
- 局部置信图
Abstract:
In visual tracking, the efficient and robust feature representation plays an important role in tracking performance in complicated environment. Therefore, a deep sparse neural network model which can extract more intrinsic and abstract features was proposed. Meanwhile, the complex and time-consuming pre-training process was avoided by using this model. During online tracking, the method of data augmentation was employed in the single positive sample to balance the quantities of positive and negative samples, which can improve the stability of the model. The local confidence maps were generated through dense sampling search to overcome the phenomenon of sampling particle drift. In order to improve the robustness of the model, several corresponding strategies of updating model parameters and searching area are proposed respectively. Extensive experimental results indicate the effectiveness and robustness of the proposed algorithm in challenging environment compared with state-of-the-art tracking algorithms. The problem of tracking drift is alleviated significantly and the tracking speed is fast.
- visual tracking /
- deep learning /
- deep sparse neural network /
- sparse autoencoders /
- local confidence maps

HTML全文

图 1 自编码器基本结构

Figure 1. Basic structure of AE

下载: 全尺寸图片幻灯片

图 2 激活函数曲线

Figure 2. Activation function curves

下载: 全尺寸图片幻灯片

图 3 跟踪网络模型

Figure 3. Model of tracking network

下载: 全尺寸图片幻灯片

图 4 对于单个正样本进行数据扩充

Figure 4. Data augmentation for single positive sample

下载: 全尺寸图片幻灯片

图 5 部分视频局部置信图示例

Figure 5. Examples of local confidence maps of some videos

下载: 全尺寸图片幻灯片

图 6 正样本时间滑动窗

Figure 6. Sliding time window of positive samples

下载: 全尺寸图片幻灯片

图 7 基于深度稀疏学习的鲁棒视觉跟踪算法流程图

Figure 7. Flowchart of robust visual tracking based on deep sparse learning

下载: 全尺寸图片幻灯片

图 8 8种跟踪算法的定性比较

Figure 8. Qualitative comparison of 8 tracking algorithms

下载: 全尺寸图片幻灯片

图 9 测试结果的精度曲线和成功率曲线

Figure 9. Precision curves and success rate curves of test results

下载: 全尺寸图片幻灯片

表 1 8种跟踪算法在10组视频中的跟踪结果对比

Table 1. Comparison of tracking results among 8 tracking algorithms on 10 videos

表 2 8种跟踪算法平均跟踪速率对比

Table 2. Comparison of average tracking speed rate among 8 tracking algorithms

跟踪算法	TLD	MIL	ASLA	CT	DFT	MTT	DLT	本文算法
平均跟踪速率/(帧·s^-1)	21.74	28.06	7.48	38.76	11.04	0.99	16.51	16.51

下载: 导出CSV

参考文献(22)

[1]	SMEULDERS A W M, CHU D M, CUCCHIARA R, et al.Visual tracking:An experimental survey[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014, 36(7):1442-1468. doi: 10.1109/TPAMI.2013.230
[2]	侯志强, 韩崇昭.视觉跟踪技术综述[J].自动化学报, 2006, 32(4):603-617. http://www.doc88.com/p-235796636077.html HOU Z Q, HAN C Z.A survey of visual tracking[J].Acta Automatica Sinica, 2006, 32(4):603-617(in Chinese). http://www.doc88.com/p-235796636077.html
[3]	LI X, HU W M, SHEN C H, et al.A survey of appearance models in visual object tracking[J].ACM Transactions on Intelligent Systems and Technology, 2013, 4(4):Article 58. http://www.oalib.com/paper/4037071
[4]	ROSS D A, LIM J, LIN R S.Incremental learning for robust visual tracking[J].International Journal of Computer Vision, 2008, 77(1-3):125-141. doi: 10.1007/s11263-007-0075-7
[5]	ZHANG T Z, GHANEM B, LIU S, et al.Robust visual tracking via multi-task sparse learning[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ:IEEE Press, 2012:2042-2049.
[6]	JIA X, LU H, YANG M H.Visual tracking via adaptive structural local sparse appearance model[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ:IEEE Press, 2012:1822-1829.
[7]	ZHANG K H, ZHANG L, YANG M H.Real-time compressive tracking[C]//Proceedings of European Conference on Computer Vision.Heidelberg:Springer Verlag, 2012, 7574:864-877.
[8]	KALAL Z, MIKOLAJCZYK K, MATAS J. Tracking-learning-detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34(7):1409-1422. doi: 10.1109/TPAMI.2011.239
[9]	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. doi: 10.1109/TPAMI.2010.226
[10]	SCHMIDHUBER J.Deep learning in neural networks:An overview[J].Neural Network, 2014, 61:85-117. http://people.idsia.ch/~juergen/deep-learning-overview.html
[11]	WANG N Y, YEUNG D.Learning a deep compact image representation for visual tracking[C]//Proceedings of Advances in Neural Information Processing Systems.Lake Tahoe:NIPS Press, 2013:809-817.
[12]	XU T Y, WU X J.Visual object tracking via deep neural network[C]//2015 IEEE 1st International Smart Cities Conference.Piscataway, NJ:IEEE Press, 2015:1-6.
[13]	ZHANG K H, LIU Q S, WU Y, et al.Robust visual tracking via convolutional networks[J].IEEE Transactions on Image Processing, 2015, 25(4):1779-1792. https://wenku.baidu.com/view/207e0d076ad97f192279168884868762caaebbc5.html
[14]	GLOROT X, BENGIO Y.Understanding the difficulty of training deep feedforward neural networks[C]//Proceedings of International Conference on Artificial Intelligence and Statistics.Brookline, MA:Microtome Publishing, 2010, 9:249-256.
[15]	WANG F S.Particle filters for visual tracking[C]//Proceedings of International Conference on Advanced Research on Computer Science and Information Engineering. Heidelberg:Springer Verlag, 2011, 152:107-112.
[16]	GLOROT X, BORDES A, BENGIO Y.Deep sparse rectifier neural networks[C]//Proceedings of International Conference on Artificial Intelligence and Statistics.Brookline, MA:Microtome Publishing, 2011, 15:315-323.
[17]	HINTON G E, SALAKHUTDINOV R.Reducing the dimensionality of data with neural networks[J].Science, 2006, 313(5786):504-507. doi: 10.1126/science.1127647
[18]	ZHANG Y, ZHANG E H, CHEN W J.Deep neural network for halftone image classification based on sparse auto-encoder[J].Engineering Applications of Artificial Intelligence, 2016, 50(1):245-255. http://www.doc88.com/p-3018954534822.html
[19]	EIGEN D, PUHRSCH C, FERGUS R.Depth map prediction from a single image using multi-scale deep network[C]//Proceedings of Advances in Neural Information Processing Systems.Montreal:Springer, 2014:2366-2374.
[20]	GAO C, CHEN F, YU J G, et al.Robust visual tracking using exemplar-based detectors[J].IEEE Transactions on Circuits & Systems for Video Technology, 2017, 27(2):300-312. http://ieeexplore.ieee.org/document/7368904/
[21]	WU Y, LIM J, YANG M H.Online object tracking:A benchmark[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ:IEEE Press, 2013, 9:2411-2418.
[22]	SEVILLA-LARA L, LEARNED-MILLER E.Distribution fields for tracking[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ:IEEE Press, 2012:1910-1917.