Xiaoyu ZHANG, Huanlong ZHANG, Xing HU, Shiqiang HU

Full occlusion handling for pedestrian tracking via hybrid system

Occlusion and lack of visibility even in sparse crowd scenes make it difficult to track individual pedestrianscorrectly and consistently, particularly in a single view. We present a novel pedestrian tracking approach that connectstracking with reidenti cation to locate and maintain the identity of certain people who may be occluded for a longtime. First, two models are constructed. One model tracks the pedestrian and trains a classi er, while the other modelreidenti es the pedestrian of interest from detection results with the trained classi er. Secondly, we design a set oftransition rules for model switching. Finally, the two models work alternatively based on the principle of a hybridsystem to track the pedestrian. Several typical sets of experiments show that the proposed approach outperforms thestate-of-the-art approaches and achieves robust pedestrian tracking in the presence of full occlusion.

PDF

___

[1]Wu Y, Lim J, Yang M. Online object tracking: a benchmark. In: IEEE 2013 Conference on Computer Vision andPattern Recognition; 25{27 June 2013; Portland, OR, USA. New York, NY, USA: IEEE. pp. 2411-2418.
[2]Yilmaz A, Javed O, Shah M. Object tracking: a survey. ACM Comput Surv 2006; 38: 1-45.
[3]Ekinci M, Gedikli E. Silhouette based human motion detection and analysis for real-time automated video surveil-lance. Turk J Elec Eng & Comp Sci 2005; 13: 199-229.
[4]Mei X, Ling H. Robust visual tracking and vehicle classi cation via sparse representation. IEEE T Pattern Anal2011; 33: 2259-2272.
[5]Zhang K, Zhang L, Yang M. Fast compressive tracking. IEEE T Pattern Anal 2014; 36: 2002-2015.
[6]Kalal Z, Mikolajczyk K, Matas J. Tracking-learning-detection. IEEE T Pattern Anal 2012; 34: 1409-1422.
[7]Papadourakis V, Argyros A. Multiple objects tracking in the presence of long-term occlusions. Comput Vis ImageUnd 2010; 114: 835-846.
[8]Yang M, Liu Y, Wen L, Zhisheng Y, Li SZ. A probabilistic framework for multitarget tracking with mutualocclusions. In: IEEE 2014 Conference on Computer Vision and Pattern Recognition; 23{28 June 2014; Columbus,OH, USA. New York, NY, USA: IEEE. pp. 1298-1305.
[9]Hua Y, Alahari K, Schmid C. Occlusion and motion reasoning for long-term tracking. In: Springer 2014 EuropeanConference on Computer Vision; 6{12 September 2014; Zurich, Switzerland. Berlin, Germany: Springer. pp. 172-187.
[10]Schaft A, Schumacher H. An Introduction to Hybrid Dynamical Systems. 1st ed. London, UK: Springer, 2000.
[11]Sankaranarayanan S, Sipma H, Manna Z. Constructing invariants for hybrid systems. In: Alur R, Pappas GJ,editors. Hybrid Systems: Computation and Control. Philadelphia, PA, USA: Springer Press, 2004. pp. 539-554.
[12]Chen S. Kalman lter for robot vision: a survey. IEEE T Ind Electron 2012; 59: 4409-4420.
[13]Diehl CP, Cauwenberghs G. SVM incremental learning, adaptation and optimization. In: IEEE 2003 Proceedingsof the International Joint Conference on Neural Networks; 20{24 July 2003; Portland, OR, USA. New York, NY,USA: IEEE. pp. 2685-2690.
[14]Zhao R, Ouyang W, Wang X. Unsupervised salience learning for person re-identi cation. In: IEEE 2013 Conferenceon Computer Vision and Pattern Recognition; 25{27 June 2013; Portland, OR, USA. New York, NY, USA: IEEE.pp. 3586-3593.
[15]Barnich O, Van Droogenbroeck M. ViBe: A universal background subtraction algorithm for video sequences. IEEET Image Process 2011; 20: 1709-1724.
[16]Ren X, Ramanan D. Histograms of sparse codes for object detection. In: IEEE 2013 Conference on ComputerVision and Pattern Recognition; 23{28 June 2013; Portland, OR, USA. New York, NY, USA: IEEE. pp. 3246-3253.
[17]Blackman SS, Popoli R. Design and Analysis of Modern Tracking Systems. Boston, MA, USA: Artech House, 1999.
[18]Bao C, Wu Y, Ling H, Ji H. Real time robust L1 tracker using accelerated proximal gradient approach. In: IEEE2012 Computer Vision and Pattern Recognition Conference; 16{21 June 2012; Rhode, Island, USA. New York, NY,USA: IEEE. pp. 1830-1837.
[19]Bo Y, Nevatia R. Multi-target tracking by online learning of non-linear motion patterns and robust appearancemodels. In: IEEE 2012 Computer Vision and Pattern Recognition Conference; 16{21 June 2012; Rhode, Island,USA. New York, NY, USA: IEEE. pp. 1918-1925.