多年來在電腦視覺的領域裡，object tracking一直是備受關注的研究議題。最近幾年，online boosting是其中一項成功應用在object tracking上的研究。當目標物移動速度沒有太快，而物體外表沒有太多appearance changes或occlusion的改變；對於這些簡單的追蹤問題，online boosting都可以有效率地即時追蹤到目標物。本論文裡，我們以online boosting的架構和概念為基礎，並且把追蹤的問題看成一種分類問題，我們將提出一個新方法，稱作：co-tracking，來進一步改善online boosting追蹤的效果。Co-tracking包含兩種意義：第一，我們考慮特徵點與特徵點之間的相互關係，藉此可以更有效地選擇出對追蹤的效果有用的特徵點；第二，我們考慮兩個不同但有關聯的目標物，例如：同時追蹤媽媽和小孩，或是一起追蹤同一人的手和腳，去分析兩者彼此的相互關係，進而改善追蹤的結果。在實驗裡，我們透過co-tracking演算法實現了穩定的追蹤效果，並且同時參考inter-feature correlation和inter-object correlation去建立更穩固的追蹤模型。我們所提出的方法能夠成功解決object tracking常遇見的問題，例如：appearance changes和occlusion；面對物體在複雜的背景裡，co-tracking依然有好的追蹤能力。

Object tracking has been one of the most focused topics in computer vi-
sion for many years. One of the most recent progress is from the success of
online boosting. The online boosting can e ectively track object in real time,
when the object moves not too fast, without too much appearance change
or occlusion; that is, online boosting can deal with most easy tracking prob-
lems. We propose a novel method to further improve the result of online
boosting for object tracking. Following the online boosting's design, we con-
sider tracking as a classi cation task. We propose the co-tracking technique
that implies two meanings: rst, we improve the tracking by considering
the correlation between di erent features, by doing that, we can more ef-
fectively select useful features for tracking; second, we improve the tracking
by considering the correlation between two di erent but related objects such
as tracking mother and son, or tracking a hand and a leg together to fur-
ther improve the tracking performance. The experiment shows that to have
considered the inter-feature correlation and inter-object correlation, we can
achieve robust tracking result. The proposed method can successfully deal
with appearance change, occlusion, and also track object when it is in front
of cluttered background.

[1] E. Alpaydin. Introduction to Machine Learning. The MIT Press, 2nd
edition, 2010.
[2] B. Babenko, M. hsuan Yang, and S. Belongie. Visual tracking with
online multiple instance learning, 2009.
[3] C. Chang, R. Ansari, and A. Khokhar. Multiple object tracking with
kernel particle lter. In Proceedings, IEEE Conference on Computer
Vision and Pattern Recognition 1 (2005) 566 573, pages 566{573, 2005.
[4] D. Comaniciu, V. Ramesh, and P. Meer. Real-time tracking of non-rigid
objects using mean shift, 2000.
[5] Y. Freund and R. E. Schapire. A decision-theoretic generalization of
on-line learning and an application to boosting, 1997.
[6] Y. Freund and R. E. Schapire. A short introduction to boosting. In In
Proceedings of the Sixteenth International Joint Conference on Arti cial
Intelligence, pages 1401{1406. Morgan Kaufmann, 1999.
[7] H. Grabner and H. Bischof. Online boosting and vision, 2006.
[8] H. Grabner, C. Leistner, and H. Bischof. Semi-supervised on-line boost-
ing for robust tracking. In Proceedings of the 10th European Conference
on Computer Vision: Part I, ECCV '08, pages 234{247, Berlin, Heidel-
berg, 2008. Springer-Verlag.
[9] M. Oren, C. Papageorgiou, P. Sinha, E. Osuna, and T. Poggio. Pedes-
trian detection using wavelet templates. In in Computer Vision and
Pattern Recognition, pages 193{199, 1997.
[10] N. C. Oza and S. Russell. Online bagging and boosting. In In Arti -
cial Intelligence and Statistics 2001, pages 105{112. Morgan Kaufmann,
2001.
[11] S. Shalev-Shwartz. Online learning: Theory, algorithms, and applica-
tions. Technical report, 2007.
[12] S. Stalder, H. Grabner, and L. V. Gool. Beyond semi-supervised track-
ing: Tracking should be as simple as detection, but not simpler than
recognition. 2009.
[13] K. Tieu and P. Viola. Boosting image retrieval. In International Journal
of Computer Vision, pages 228{235, 2000.
[14] P. Viola and M. Jones. Rapid object detection using a boosted cascade
of simple features. pages 511{518, 2001.
[15] P. Viola and M. Jones. Robust real-time face detection. International
Journal of Computer Vision, 57:137{154, 2004.
[16] A. Yilmaz, O. Javed, and M. Shah. Object tracking: A survey. ACM
Comput. Surv., 38(4), Dec. 2006.
[17] Q. Yu, T. B. Dinh, and G. Medioni. Online tracking and reacquisition
using co-trained generative and discriminative trackers. In Proceedings
of the 10th European Conference on Computer Vision: Part II, ECCV
'08, pages 678{691, Berlin, Heidelberg, 2008. Springer-Verlag.