Open AccessPerson re‐identification with discriminatively trained viewpoint invariant orthogonal dictionaries
Author(s) -
Gao Bin,
Zeng Mingyong,
Xu Shiming,
Sun Fenggang,
Guo Jibin
Publication year - 2016
Publication title -
electronics letters
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.375
H-Index - 146
ISSN - 1350-911X
DOI - 10.1049/el.2016.2639
Subject(s) - invariant (physics) , artificial intelligence , computer science , pattern recognition (psychology) , identification (biology) , speech recognition , mathematics , botany , mathematical physics , biology
A novel and efficient method for person re‐identification based on orthogonal dictionary learning is proposed. The orthogonal dictionary exhibits extraordinary discriminative power than the classical dictionary learning. It is learned with the help of convex optimisation and customised trace optimisation. The approach has been evaluated against current methods on a benchmark dataset and can reach outstanding performance.