Robust and Fast Tracking-Learning-Detection

DOI

10.2991/csic-15.2015.109How to use a DOI?

Keywords

Tracking-Learning-Detection, Failure predictors, RANSAC, Weighted P-N learning

Abstract

To improve the robust and processing speed of the Tracking-Learning-Detection(TLD), the robust and fast TLD tracker is proposed. Replacing with the forward-backward error predictor, The two powerful failure predictors, including the neighbourhood consistency predictor and the markov predictor in the tracker, are used to reduce the computational cost and improve the precise. The RANSAC algorithm is added to estimate the global motion model and improve the success rate of tracking. Replacing with P-N learning in sample learning procedure, We use a novel online weighted P-N learning which integrates the sample importance into an efficient online learning procedure to alleviate drift to some extent. Experimental results on various benchmark video sequences demonstrate the superior performance of the proposed algorithm to state-of-the-art tracking algorithms in robustness, stability and efficiency.

Copyright

© 2015, 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  - Shuai Cheng
AU  - Guangwen Liu
AU  - Junxi Sun
PY  - 2015/07
DA  - 2015/07
TI  - Robust and Fast Tracking-Learning-Detection
BT  - Proceedings of the 2015 International Conference on Computer Science and Intelligent Communication
PB  - Atlantis Press
SP  - 449
EP  - 452
SN  - 2352-538X
UR  - https://doi.org/10.2991/csic-15.2015.109
DO  - 10.2991/csic-15.2015.109
ID  - Cheng2015/07
ER  -