A Novel Two-Stage PCA Algorithm for Object Tracking

DOI

10.2991/cimns-16.2016.72How to use a DOI?

Keywords

particle filter; two-stage PCA; eigenspace; reconstruction error

Abstract

In this paper, we propose a two-stage PCA algorithm to deal with the problem of target appearance changes in object tracking. Our method is based on particle filter framework and aims at building up a robust appearance model for the target. In the first-stage, PCA is applied on several templates collected before tracking to construct a low dimensional subspace for the targets. Correspondingly, in the later tracking, all particles will be projected into this subspace to calculate the weight of particles, as well as the location of the target by weighted sum of particles. In the second-stage, PCA is adopted to decompose the reconstruction error into orthogonal basis to find the basis with biggest variance which can best present the appearance changes. Correspondingly, we uses this basis to update the old subspace. Besides, a threshold is set to decide when to update which greatly reduce the numbers of update. The two stages work together to establish a robust appearance model making our tracking algorithm more robust. Experimental results on public video sequence demonstrate the effectiveness of our proposed algorithm.

Copyright

© 2016, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Yuanyuan Yang
AU  - Dong Hu
PY  - 2016/09
DA  - 2016/09
TI  - A Novel Two-Stage PCA Algorithm for Object Tracking
BT  - Proceedings of the 2016 International Conference on Communications, Information Management and Network Security
PB  - Atlantis Press
SP  - 289
EP  - 292
SN  - 2352-538X
UR  - https://doi.org/10.2991/cimns-16.2016.72
DO  - 10.2991/cimns-16.2016.72
ID  - Yang2016/09
ER  -