KL based data fusion for target tracking

Jing Peng, K. Palaniappan, Sema Candemir, Guna Seetharaman

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Visual object tracking in video can be formulated as a time varying appearance-based binary classification problem. Tracking algorithms need to adapt to changes in both foreground object appearance as well as varying scene backgrounds. Fusing information from multimodal features (views or representations) typically enhances classification performance without increasing classifier complexity when image features are concatenated to form a high-dimensional vector. Combining these representative views to effectively exploit multimodal information for classification becomes a key issue. We show that the Kullback-Leibler (KL) divergence measure provides a framework that leads to family of techniques for fusing representations including Cher-noff distance and variance ratio that is the same as linear discriminant analysis. We provide experimental results that corroborate well with our theoretical analysis.

Original languageEnglish
Title of host publicationICPR 2012 - 21st International Conference on Pattern Recognition
Pages3480-3483
Number of pages4
StatePublished - 1 Dec 2012
Event21st International Conference on Pattern Recognition, ICPR 2012 - Tsukuba, Japan
Duration: 11 Nov 201215 Nov 2012

Publication series

NameProceedings - International Conference on Pattern Recognition
ISSN (Print)1051-4651

Other

Other21st International Conference on Pattern Recognition, ICPR 2012
CountryJapan
CityTsukuba
Period11/11/1215/11/12

Fingerprint Dive into the research topics of 'KL based data fusion for target tracking'. Together they form a unique fingerprint.

Cite this