基于代价敏感结构化SVM的目标跟踪

袁广林; 孙子文; 秦晓燕; 夏良; 朱虹

doi:10.11999/JEIT200708

基于代价敏感结构化SVM的目标跟踪

doi: 10.11999/JEIT200708

中国人民解放军陆军炮兵防空兵学院信息工程系合肥 230031

基金项目: 安徽省自然科学基金(2008085QF325)

详细信息

作者简介:
袁广林：男，1973年生，博士，副教授，主要研究方向为图像处理、计算机视觉、机器学习及其应用等

孙子文：男，1996年生，硕士生，研究方向为计算机视觉、机器学习

秦晓燕：女，1980年生，硕士，讲师，主要研究方向为计算机视觉、机器学习等

夏良：男，1980年生，硕士，副教授，主要研究方向为计算机视觉、机器学习、大数据等

朱虹：女，1987年生，硕士，讲师，主要研究方向为计算机视觉、图像处理等

通讯作者:
袁广林　yuangl_plus@126.com

中图分类号: TN911.73; TP391.4
计量
- 文章访问数: 877
- HTML全文浏览量: 459
- PDF下载量: 83
- 被引次数: 0
出版历程
- 收稿日期: 2020-08-10
- 修回日期: 2021-04-14
- 网络出版日期: 2021-07-11
- 刊出日期: 2021-11-23

Object Tracking Based on Cost Sensitive Structured SVM

Department of Information Engineering of PLA Army Academy of Artillery and Air Defense, Hefei 230031, China

Funds: Anhui Provincial Natural Science Foundation (2008085QF325)

摘要

摘要: 基于结构化SVM的目标跟踪由于其优异的性能而受到了广泛关注，但是现有方法存在正样本和负样本不平衡问题。针对此问题，该文首先提出一种用于目标跟踪的代价敏感结构化SVM模型，其次基于对偶坐标下降原理设计了该模型的求解算法，最后利用提出的代价敏感结构化SVM实现了一种多尺度目标跟踪方法。在OTB100数据集和VOT2019数据集上进行了实验验证，实验结果表明：该文方法相比相关滤波目标跟踪方法，跟踪精度较高，相比深度目标跟踪方法，具有速度优势。
- 目标跟踪 /
- 非平衡问题 /
- 代价敏感 /
- 结构化SVM
Abstract: Object tracking based on structured SVM attracts much attention due to its excellent performance. However, the existing methods have the problem of imbalance between positive and negative samples. To solve the problem, a cost sensitive structured SVM model is proposed for object tracking. Secondly, an algorithm for the proposed model is designed via dual coordinate descent principle. Finally, a multi-scale object tracking method is implemented using the proposed cost sensitive structured SVM. The experimental results on OTB100 datasets and VOT2019 datasets show that compared with the correlation filtering trackers, the proposed method has higher tracking accuracy, and has the advantage of speed compared with the deep object trackers.
- Object tracking /
- Unbalanced problem /
- Cost sensitive /
- Structured SVM

HTML全文

图 1 结构化SVM目标跟踪中存在的正样本和负样本不平衡问题

下载: 全尺寸图片幻灯片

图 2 6种高性能跟踪器在OTB100数据集上取效果前5名的OPE, TRE和SRE性能指标曲线

下载: 全尺寸图片幻灯片

表 1 6种基于结构化SVM的跟踪器在OTB100数据集上的OPE性能与速度指标

跟踪方法	精度(pixels)	成功率(AUC)	速度(fps)
Scale-DLCS_SSVM	0.837	0.648	12.18
DLCS_SSVM	0.805	0.603	25.32
Scale-DLSSVM^[8]	0.803	0.561	11.47
DLSSVM^[8]	0.765	0.540	23.16
Struck^[1]	0.635	0.459	7.52
LMCF^[9]	0.786	0.578	78.23

下载: 导出CSV

表 2 5种跟踪方法在OTB100数据集上的速度指标(fps)

跟踪方法	Scale-DLCS_SSVM	DLCS_SSVM	DeepLMCF^[9]	DeepSRDCF^[17]	TADT^[18]
速度	12.18	25.32	7.45	1.56	27.89

下载: 导出CSV

表 3 本文多尺度方法与近年来4种高性能跟踪方法在VOT2019数据集上实验结果

	本文	SiamMask^[20]	GradNet^[21]	TADT^[18]	DeepSRDCF^[17]
EAO↑	0.217	0.289	0.213	0.207	0.198
Accuracy↑	0.538	0.596	0.531	0.516	0.512
Robustness↓	0.559	0.461	0.538	0.567	0.572

下载: 导出CSV

参考文献(21)

[1]	HARE S, SAFFARI A, and TORR P H S. Struck: Structured output tracking with kernels[C]. 2011 International Conference on Computer Vision, Barcelona, Spain, 2011: 263−270.
[2]	HARE S, GOLODETZ S, SAFFARI A, et al. Struck: Structured output tracking with kernels[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(10): 2096–2109. doi: 10.1109/TPAMI.2015.2509974
[3]	YAO Rui, SHI Qinfeng, SHEN Chunhua, et al. Robust tracking with weighted online structured learning[C]. The 12th European Conference on Computer Vision, Florence, Italy, 2012: 158–172.
[4]	YAO Rui, SHI Qinfeng, SHEN Chunhua, et al. Part-based visual tracking with online latent structural learning[C]. 2013 IEEE Conference on Computer Vision and Pattern Recognition, Portland, USA, 2013: 2363–2370.
[5]	BAI Yancheng and TANG Ming. Robust tracking via weakly supervised ranking SVM[C]. 2012 IEEE Conference on Computer Vision and Pattern Recognition, Providence, USA, 2012: 1854–1861.
[6]	ZHANG Jianming, MA Shugao, and SCLAROFF S. MEEM: Robust tracking via multiple experts using entropy minimization[C]. The 13th European Conference on Computer Vision, Zurich, Swiss, 2014: 188–203.
[7]	HONG S, YOU T, KWAK S, et al. Online tracking by learning discriminative saliency map with convolutional neural network[C]. The 32nd International Conference on International Conference on Machine Learnin, Lille, France, 2015: 597–606.
[8]	NING Jifeng, YANG Jimei, JIANG Shaojie, et al. Object tracking via dual linear structured SVM and explicit feature map[C]. 2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, USA, 2016: 4266–4274.
[9]	WANG Mengmeng, LIU Yong, and HUANG Zeyi. Large margin object tracking with circulant feature maps[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, USA, 2017: 4800–4808.
[10]	JI Zhangjian, FENG Kai, and QIAN Yuhua. Part-based visual tracking via structural support correlation filter[J]. Journal of Visual Communication and Image Representation, 2019, 64: 102602. doi: 10.1016/j.jvcir.2019.102602
[11]	ZUO Wangmeng, WU Xiaohe, LIN Liang, et al. Learning support correlation filters for visual tracking[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 41(5): 1158–1172. doi: 10.1109/TPAMI.2018.2829180
[12]	ZABIH R and WOODFILL J. Non-parametric local transforms for computing visual correspondence[C]. The Third European Conference on Computer Vision, Stockholm, Sweden, 1994: 151–158.
[13]	WU Yi, LIM J, and YANG M H. Object tracking benchmark[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1834–1848. doi: 10.1109/TPAMI.2014.2388226
[14]	KRISTAN M, BERG J, LEONARDIS A, et al. The seventh visual object tracking VOT2019 challenge results[C]. 2019 IEEE/CVF International Conference on Computer Vision Workshop, Seoul, South Korea, 2019: 2206–2241.
[15]	IRANMEHR A, MASNADI-SHIRAZI H, and VASCONCELOS N. Cost-sensitive support vector machines[J]. Neurocomputing, 2019, 343: 50–64. doi: 10.1016/j.neucom.2018.11.099
[16]	RAMANAN D. Dual coordinate solvers for large-scale structural SVMs[J]. arXiv: 1312.1743v2, 2014.
[17]	DANELLJAN M, HÄGER G, KHAN F S, et al. Convolutional features for correlation filter based visual tracking[C]. 2015 IEEE International Conference on Computer Vision Workshop, Santiago, Chile, 2015: 621–629.
[18]	WANG Qiang, ZHANG Li, BERTINETTO L, et al. Fast online object tracking and segmentation: A unifying approach[C]. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, USA, 2019: 1328–1338.
[19]	LI Peixia, CHEN Boyu, OUYANG Wanli, et al. GradNet: Gradient-guided network for visual object tracking[C]. 2019 IEEE/CVF International Conference on Computer Vision, Seoul, South Korea, 2019: 6161–6170.
[20]	LI Xin, MA Chao, WU Baoyuan, et al. Target-aware deep tracking[C]. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, USA, 2019: 1369–1378.
[21]	BERTINETTO L, VALMADRE J, GOLODETZ S, et al. Staple: Complementary learners for real-time tracking[C]. 2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, USA, 2016: 1401–1409.