基于带权重多样例学习的视觉跟踪算法

毛征; 贾文洋; 杜文彬; 梅伟军

doi:10.11936/bjutxb2016030069

基于带权重多样例学习的视觉跟踪算法

doi: 10.11936/bjutxb2016030069

北京工业大学电子信息与控制工程学院, 北京 100124

基金项目: 国家自然科学基金资助项目(81370038)

详细信息

中图分类号: TP391.9
计量
- 文章访问数: 83
- HTML全文浏览量: 65
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2016-03-28
- 网络出版日期: 2022-09-13
- 刊出日期: 2017-02-01

Visual Tracking Method Based on Weighted Sample Learning

College of Electronic and control Engineering, Beijing University of Technology, Beijing 100124

摘要

摘要: 原始在线加权的多示例学习跟踪假设每个示例是独立且在包中的贡献均相同, 同时为所有正样本赋予相同的权重,这不符合“包中的示例与目标位置的远近,对目标贡献程度是不一样”的事实. 再加上原始算法采取单一特征无法准确和全面地表示目标包中所包含的示例,从而影响了跟踪算法的鲁棒性. 针对原始算法的这些问题,提出一种基于带权重多样例学习的视觉跟踪方法. 该方法同时融合多特征(HOG特征和Haar特征),在多示例学习框架下同时训练分类器,并通过样本特征相似度的比较来赋予不同的权重. 对不同场景的图像序列进行实验,通过在公共测试集上与多种主流算法做对比,显示这样得到的目标外表模型对于前景和背景具有更高的区分能力. 结果表明:该算法具有更高的准确性和更强的适应性,可以有效克服传统多示例学习中的分类器退化问题.
- 多样例学习 /
- 多特征 /
- 视觉跟踪
Abstract: Original online weighted sample learning tracking assumes that each sample is more independent and its contribution to the package is all the same, and the same weight is given for all the positive samples. Therefore, it does not agree with the fact that the contribution degree of the target is not all the same with the distance between the target position and the sample of package. Additionally, the original algorithm cannot accurately and comprehensively represent the sample of the target package because of single feature, thus, it affects the robustness of the algorithm. Regarding the problem of original algorithm, this paper put forward a type of visual tracking method based on weighted sample learning. The method fused HOG features and Haar features at the same time, trained a classifier under the framework of learning, and gave different weights according to the similarity of the sample features. The experiments were conducted on the image sequence of different scenarios, and compared with a variety of current mainstream algorithms. Results show that the created appearance model has higher ability to distinguish between foreground and background, the algorithm has higher accuracy and stronger adaptability and can effectively overcome the traditional sample classifier degradation problems in learning.
- multiple instance learning /
- more features /
- visual tracking
The authors have declared that no competing interests exist.

HTML全文

图 1 训练样本采样流程

Figure 1. Sampling procedure of training samples

下载: 全尺寸图片幻灯片

图 2 候选样本采样

Figure 2. Sampling procedure of candidate samples

下载: 全尺寸图片幻灯片

图 3 基于特征相似度的采样

Figure 3. Sampling based on feature similarity

下载: 全尺寸图片幻灯片

图 4 各个算法在tiger上的跟踪结果

Figure 4. Tracking results of different algorithms on tiger

下载: 全尺寸图片幻灯片

图 5 各个算法在sylv上的跟踪结果

Figure 5. Tracking results of different algorithms on sylv

下载: 全尺寸图片幻灯片

图 6 各个算法在facocc上的跟踪结果

Figure 6. Tracking results of different algorithms on facocc

下载: 全尺寸图片幻灯片

参考文献(14)

[1]	BABENKO B, YANG M H, BELONGIE S.Robust object tracking with online multiple instance learning[J]. IEEE Transactions on Patten Analysis and Machine Intelligence, 2011, 33(8): 1619-1632.
[2]	ZHANG K, SONG H.Real-time visual tracking via online weighted multiple instance learning[J]. Pattern Recognition, 2013, 46(1): 397-411.
[3]	PENG S, PENG X M.Object tracking with efficient multiple instance learning[J]. Journal of Computer Applications, 2015, 35(2) : 466-469. (in Chinese)
[4]	MAO Z, YUAN J J, WU Z R, et al.Real-time compressive tracking based on online feature selection[C]//Proceedings of International Conference on Computer Science and Information Technology. Berlin: Springer, 2014: 431-438.
[5]	ZHU Q P, YAN J, ZHANG H, et al.Real-time tracking using multiple features based on compressive sensing[J]. Optics and Precision Engineering, 2013, 21(2): 437-444. (in Chinese)
[6]	GAO Y.Based on the characteristics of Haar human movement detection [D]. Changchun: Jilin University, 2013. (in Chinese)
[7]	GEISMANNPSCHNEIDERG.A two-staged approach to vision-based pedestrian recognition using Haar and HOG featureIntelligent Vehicles Symposium. Eindhoven: IEEE2008554559 GEISMANN P, SCHNEIDER G.A two-staged approach to vision-based pedestrian recognition using Haar and HOG feature[C]//Intelligent Vehicles Symposium. Eindhoven: IEEE, 2008: 554-559.
[8]	LU H, ZHOU Q, WANG D, et al.A co-training framework for visual tracking with multiple instance learning[C]//IEEE International Conference on Automatic Face and Gesture Recognition. New York: IEEE, 2011: 539-544.
[9]	HUANG R J, LI Y, LI W H, et al.AdaBoost for pedestrian detection based on multi-feature[J]. Journal of Jilin University, 2010, 13(3): 449-455. (in Chinese)
[10]	LI W B, LIU C N, CHEN N Y.Text classification algorithm based on the weight of feature information[J]. Journal of Beijing University of Technology, 2006, 32(5): 456-460. (in Chinese)
[11]	ZHOU Z H, YANG Y, WU X D, et al.The top ten algorithms in data mining[M]. New York: CRC Press, 2009, 127-149.
[12]	YNAG D L, ZHAO Z M, BAI Q C.Face detection based on AdBoost and space support vector domain classifier[J]. Journal of Circuits and Systems, 2013, 39(6): 321-325. (in Chinese)
[13]	CHEN D C, ZHU M, GAO W, et al.Real-time object tracking via online weighted multiple instance learning[J]. Optics and Precision Engineering, 2014, 22(6): 1661-1667. (in Chinese)
[14]	CHEN C S, ZHANG J, XIE Z, et al.Multi-cues object tracking based on motion consistence in random field[J]. Journal of Image and Graphics, 2015, 20(1) : 59-71. (in Chinese)