物件追蹤(Object tracking) 是近年被廣為研究的影像處理問題，主要在影片
中持續擷取目標物件的特徵，它可以被應用在影片編輯、影片監視、影片復原和
影片壓縮...... 等。但在追蹤目標物件中，有時會因為頻繁遮蔽、物件消失、類似
目標的其他物件、偵測失誤和亮度變化而導致資訊的丟失。而我們的做法可以藉
由偵測輸入幀(frame)，判斷出追蹤的目標物件位置，並且整合學習和追蹤系統，
使系統可以即時更新，避免上述問題干擾。我們假設影片第一幀能獲得初始的目
標物件和位置，經由Structured output SVM 建立我們的初始訓練模型。接著追
蹤每一幀時，我們利用前一幀的位置計算搜尋範圍內是否有目標物件，當偵測
到目標物件後將其鎖定。並且計算追蹤框尺度的轉換，先運用SIFT+RANSAC
將矩形框架適應目標物件，接著在框架中做過度的區域分割(oversegmentation)。
將分割出來的所有子區域關係建立成無向圖(undirected graph) 後，為了計算
框架中擁有目標物件的區域，我們要尋找分數最高的連續集合，所以轉換成
Prize-collecting Steiner Tree (PCST) 問題。找出最高分數的區域集合，校正目標
物件框架後，更新訓練模型和搜尋範圍的中心位置。而我們的實驗結果和近年來
的方法比較，經過評估後，大部分結果勝於其他方法。

Object tracking has been studied broadly as image processing issue for the past
few years and the main purpose continually captures the object’s character. It can be
applied to video editing, video surveillance, video compression, video retrieval, and
etc. But when tracking the objecting, sometimes we lose the object’s information
due to frequent occlusions, disappeared object, similar target appearances, missed
detection and illumination change. We provide a system to directly predict the next
frame’s position with changing and immediately refresh the system by combining
learn with track. It defines that the first frame of video has original object and
position and builds the original tracking model according to structured output SVM.
To track every frame, system uses the last position to calculate and track range
which if exists object or not. After tracking the object, and transferring the scale.
System uses SIFT+RANSAC to match between rectangular window and object
before oversegmentation of rectangular window. After building all of the segmented
sub regions to undirected graph, we have to find out the continuous set of the bestscore
in order to calculate the area having target object in the rectangular window.
Therefore, we turn the issue into Prize-collecting Steiner Tree (PCST) and find out
the continuous set of the best-score and aims the rectangular window of object to
refresh structured output SVM tracking model and frame position. After estimating,
the experimental data compared to the recent methods is better than others.

[1] S. Hare, A. Saffari, and P. H. Torr, “Struck: Structured output tracking
with kernels,” in IEEE International Conference on Computer Vision (ICCV),
pp. 263–270, 2011.
[2] Z. Kalal, K. Mikolajczyk, and J. Matas, “Tracking-learning-detection,” in IEEE
Transactions on Pattern Analysis and Machine Intelligence(TPAMI), vol. 34,
pp. 1409–1422, 2012.
[3] K. Zhang, L. Zhang, and M.-H. Yang, “Real-time compressive tracking,” in
European Conference on Computer Vision(ECCV), vol. 7574, pp. 864–877,
Springer Berlin Heidelberg, 2012.
[4] T. B. Dinh, N. Vo, and G. Medioni, “Context tracker: Exploring supporters and
distracters in unconstrained environments,” in IEEE International Conference
on Computer Vision and Pattern Recognition (CVPR), pp. 1177–1184, 2011.
[5] A. Yilmaz, O. Javed, and M. Shah, “Object tracking: A survey,” in ACM
Computing Surveys (CSUR), vol. 38, 2006.
[6] M. A. Zulkifley, B. Moran, and D. Rawlinson, “Robust hierarchical multiple
hypothesis tracker for multiple object tracking,” in IEEE International Conference
on Image Processing (ICIP), pp. 405–408, 2012.
[7] S. Avidan, “Support vector tracking,” in IEEE Transactions on Pattern Analysis
and Machine Intelligence(TPAMI), vol. 26, pp. 1064–1072, 2004.
[8] Z. Kalal, J. Matas, and K. Mikolajczyk, “P-n learning: Bootstrapping binary
classifiers by structural constraints,” in IEEE International Conference
on Computer Vision and Pattern Recognition (CVPR), pp. 49–56, 2010.
[9] A. Bordes, L. Bottou, P. Gallinari, and J. Weston, “Solving multiclass support
vector machines with larank,” in ACM International Conference on Machine
learning(ICML), pp. 89–96, 2007.
[10] Y. Kang, W. Xie, and B. Hu, “A scale adaptive mean-shift tracking algorithm
for robot vision,” in Journal of Advances in Mechanical Engineering, vol. 2013,
Hindawi Publishing Corporation, 2013.
[11] M.-Y. Liu, O. Tuzel, S. Ramalingam, and R. Chellappa, “Entropy rate superpixel
segmentation,” in IEEE International Conference on Computer Vision
and Pattern Recognition (CVPR), pp. 2097–2104, 2011.
[12] S. Vijayanarasimhan and K. Grauman, “Efficient region search for object detection,”
in IEEE International Conference on Computer Vision and Pattern
Recognition (CVPR), pp. 1401–1408, 2011.
[13] I. Ljubić, R. Weiskircher, U. Pferschy, G. W. Klau, P. Mutzel, and M. Fischetti,
“An algorithmic framework for the exact solution of the prize-collecting steiner
tree problem,” in Mathematical Programming, vol. 105, pp. 427–449, Springer
Berlin Heidelberg, 2006.
[14] C. H. Lampert, M. B. Blaschko, and T. Hofmann, “Beyond sliding windows:
Object localization by efficient subwindow search,” in IEEE International Conference
on Computer Vision and Pattern Recognition (CVPR), pp. 1–8, 2008.
[15] A. Lehmann, B. Leibe, and L. Van Gool, “Feature-centric efficient subwindow
search,” in IEEE International Conference on Computer Vision(ICCV),
pp. 940–947, 2009.
[16] T. Yeh, J. J. Lee, and T. Darrell, “Fast concurrent object localization and
recognition,” in IEEE International Conference on Computer Vision and Pattern
Recognition (CVPR), pp. 280–287, 2009.
[17] W. Zhu, S. Wang, R.-S. Lin, and S. Levinson, “Tracking of object with svm
regression,” in IEEE International Conference on Computer Vision and Pattern
Recognition (CVPR), vol. 2, pp. II–240, 2001.
[18] I. Tsochantaridis, T. Hofmann, T. Joachims, and Y. Altun, “Support vector
machine learning for interdependent and structured output spaces,” in ACM
International Conference on Machine learning(ICML), p. 104, 2004.
[19] A. Bordes, N. Usunier, and L. Bottou, “Sequence labelling svms trained in one
pass,” in Machine Learning and Knowledge Discovery in Databases, pp. 146–
161, Springer Berlin Heidelberg, 2008.
[20] J. Platt et al., “Sequential minimal optimization: A fast algorithm for training
support vector machines,” in Microsoft Research technical report msr-tr-98-14,
1998.
[21] X. Wang, C. Ning, A. Shi, and G. Lv, “An improved similarity measure in
particle filters for robust object tracking,” in IEEE International Congress on
Image and Signal Processing (CISP), vol. 1, pp. 46–50, 2013.
[22] G. Shi, X. Xu, and Y. Dai, “Sift feature point matching based on improved
ransac algorithm,” in IEEE International Conference on Intelligent Human-
Machine Systems and Cybernetics (IHMSC), vol. 1, pp. 474–477, 2013.
[23] D. G. Lowe, “Object recognition from local scale-invariant features,” in IEEE
International Conference on Computer Vision(ICCV), vol. 2, pp. 1150–1157,
1999.
[24] K. Mikolajczyk and C. Schmid, “An affine invariant interest point detector,”
in European Conference on Computer Vision(ECCV), pp. 128–142, Springer
Berlin Heidelberg, 2002.
[25] K. Levenberg, “A method for the solution of certain non-linear problems in
least squares,” in Quarterly Journal of Applied Mathmatics, vol. II, pp. 164–
168, 1944.
[26] I. Ljubić, R. Weiskircher, U. Pferschy, G. W. Klau, P. Mutzel, and M. Fischetti,
“An algorithmic framework for the exact solution of the prize-collecting steiner
tree problem,” in Mathematical Programming, vol. 105, pp. 427–449, Springer
Berlin Heidelberg, 2006.
[27] J. Kwon and K. M. Lee, “Tracking of a non-rigid object via patch-based dynamic
appearance modeling and adaptive basin hopping monte carlo sampling,” in
IEEE International Conference on Computer Vision and Pattern Recognition
(CVPR), pp. 1208–1215, 2009.
[28] D. Wang, H. Lu, and M.-H. Yang, “Online object tracking with sparse prototypes,”
in IEEE Transaction on Image Processing(TIP), vol. 22, pp. 314–325,
2013.
[29] Y. Wu, J. Lim, and M.-H. Yang, “Online object tracking: A benchmark,” in
IEEE International Conference on Computer Vision and Pattern Recognition
(CVPR), pp. 2411–2418, 2013.
[30] F. Bashir and F. Porikli, “Performance evaluation of object detection and tracking
systems,” in IEEE International Workshop on Performance Evaluation of
Tracking and Surveillance (PETS), vol. 5, 2006.