Title :
Comparison of discriminant analysis methods applied to diffractive optically variable image
Author :
Hu, Rukun ; Feng, Youji ; Guo, Ping
Author_Institution :
Image Process. & Pattern Recognition Lab., Beijing Normal Univ., Beijing
Abstract :
As a kind of powerful anti-counterfeiting device, diffractive optically variable image (DOVI) has been developed and widely used in information security field. However, the identification of DOVI today by bare eyes is not reliable. In this paper we investigate the recognition of DOVI with machine learning method, and five kinds of algorithms, namely quadratic discriminate analysis (QDA), linear discriminate analysis (LDA), regularized discriminate analysis (RDA), leave-one-out covariance matrix estimate (LOOC), and Kullback-Leibler information measure based method (KLIM) are applied to the recognition of DOVI. Considering both time cost and correct classification rate, KLIM classifier exceeds others.
Keywords :
image classification; image recognition; learning (artificial intelligence); security of data; Kullback-Leibler information; anticounterfeiting device; classification rate; diffractive optically variable image; discriminant analysis methods; information security; leave-one-out covariance matrix estimate; linear discriminate analysis; machine learning; quadratic discriminate analysis; regularized discriminate analysis; time cost; Algorithm design and analysis; Eyes; Image analysis; Information analysis; Information security; Learning systems; Linear discriminant analysis; Machine learning algorithms; Optical devices; Optical diffraction; Diffractive optically variable image; Discriminant analysis; Pattern recognition; Regularization;
Conference_Titel :
Anti-counterfeiting, Security and Identification, 2008. ASID 2008. 2nd International Conference on
Conference_Location :
Guiyang
Print_ISBN :
978-1-4244-2584-6
Electronic_ISBN :
978-1-4244-2585-3
DOI :
10.1109/IWASID.2008.4688337
