Compressive tracking via appearance modeling based on structural local patch

In this paper, we propose a compressive tracking method via appearance model based on structural local patchs and improved Haar-like feature. In contrast to previous compressive tracking only considering the holistic representation, an object can be represented by local image patches with spatial layout in an object. This representation takes advantage of both partial information and spatial information of the target. Each local patch has a fixed position in the target field, and all local patches can represent the whole target. In addition, our appearance model based on features extracted from image patches, which can guarantee the randomness of the rectangular boxes and the distribution of the rectangular boxes over the entire image area, avoiding the randomness of the rectangular boxes is too strong to weak the feature expression. We sample the positive and negative samples and divide them into patchs to train a binary classification via a naive Bayes classifier with online update, then the classifier is used to discriminate the candidate samples. The candidate sample which gets the highest classify score is the target. After that we draw positive and negative samples in the same way with the candidate samples to update the classifier to get ready for next frame. Our approach helps not only locate the target more accurately but also can handle partial occlusion effectively. The proposed tracker is compared with several state-of-the-art trackers on some challenging video sequences. Our proposed tracker is better and more stable in both quantitative and qualitative comparisons.

[1]  Ming-Hsuan Yang,et al.  Incremental Learning for Robust Visual Tracking , 2008, International Journal of Computer Vision.

[2]  Dimitris Achlioptas,et al.  Database-friendly random projections: Johnson-Lindenstrauss with binary coins , 2003, J. Comput. Syst. Sci..

[3]  Junseok Kwon,et al.  Visual tracking decomposition , 2010, 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[4]  Kenneth Ward Church,et al.  Very sparse random projections , 2006, KDD '06.

[5]  E. Candès,et al.  Stable signal recovery from incomplete and inaccurate measurements , 2005, math/0503066.

[6]  Huchuan Lu,et al.  Visual tracking via adaptive structural local sparse appearance model , 2012, 2012 IEEE Conference on Computer Vision and Pattern Recognition.

[7]  R. DeVore,et al.  A Simple Proof of the Restricted Isometry Property for Random Matrices , 2008 .

[8]  Lei Zhang,et al.  Real-Time Compressive Tracking , 2012, ECCV.

[9]  Junseok Kwon,et al.  Tracking of a non-rigid object via patch-based dynamic appearance modeling and adaptive Basin Hopping Monte Carlo sampling , 2009, CVPR.

[10]  Takahiro Ishikawa,et al.  The template update problem , 2004, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[11]  Ming-Hsuan Yang,et al.  Visual tracking with online Multiple Instance Learning , 2009, CVPR.

[12]  Huchuan Lu,et al.  Superpixel tracking , 2011, 2011 International Conference on Computer Vision.