COMPRESSIVE SENSING WITH WEIGHTED LOCAL CLASSIFIERS FOR ROBOT VISUAL TRACKING
Wenhui Huang, Jason Gu, and Xin Ma
Keywords
Compressive tracking, weighted local classifiers, particle filter, robot visual tracking
Abstract
In this paper, we propose an improved compressive tracking (CT) algorithm with weighted local classifiers (WLCs) under the particle filter framework for robot visual tracking. WLCs address some drift problems caused by partial interference, such as partial occlusion or partial appearance variation occurring under certain circumstances,
by dividing each interest candidate into several rectangular sub-regions (or local regions), each of which is independently trained with a local naive Bayesian classifier. Therefore, the object location is considered the candidate with the maximal response of a global
classifier, which is a combination of the WLCs. Furthermore, we incorporate the particle filter into CT to account for the motion of the object and to better estimate the object location. The experimental results demonstrate the effectiveness and robustness of
our proposed tracking algorithm.
Important Links:
Go Back