Open AccessMethods for accuracy‐preserving acceleration of large‐scale comparisons in CPU‐based iris recognition systems
Author(s) -
Rathgeb Christian,
Buchmann Nicolas,
Hofbauer Heinz,
Baier Harald,
Uhl Andreas,
Busch Christoph
Publication year - 2018
Publication title -
iet biometrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.434
H-Index - 28
ISSN - 2047-4946
DOI - 10.1049/iet-bmt.2016.0125
Subject(s) - iris recognition , computer science , biometrics , artificial intelligence , iris (biosensor) , identification (biology) , hamming distance , software , toolbox , pattern recognition (psychology) , computer vision , data mining , botany , algorithm , biology , programming language
To confirm an individual's identity accurately and reliably iris recognition systems analyse the texture that is visible in the iris of the eye. The rich random pattern of the iris constitutes a powerful biometric characteristic suitable for biometric identification in large‐scale deployments. Identification attempts or deduplication checks require an exhaustive one‐to‐many comparison. Hence, for large‐scale biometric databases with millions of enrollees, the time required for a biometric identification is expected to significantly increase. In this study, the authors analyse techniques to accelerate Hamming distance‐based comparisons of binary biometric reference data, i.e. iris‐codes, in large‐scale iris recognition systems, which preserve the biometric performance. The focus is put on software‐based optimisations, an efficient two‐step iris‐code alignment process referred to as TripleA , and a combination thereof. Benchmarking the throughput and identifying potential bottlenecks of a portable commodity hardware‐based iris recognition system is of particular interest. Based on the conducted experiments the authors point out practical boundaries of large‐scale comparisons in central processing unit‐based iris recognition systems, bridging the gap between the fields of iris recognition and software design.