Feature extraction using FFT for Banknotes Recognition in a variety of lighting conditions

The watermarks are not only imprinted on the bank notes so as to recognize the nationality as well as its denomination, but also respond to a variety of illuminations such as Ultraviolet, Infrared and white ray easily. In this paper, we propose a discrimination algorithm for the nationality of banknote by using SIFT and FFT features. Lots of features of the banknotes are extracted under an illumination while varying the type of illuminations, which are used as a data base for the Multinational Banknotes Recognition Engine [MBRE]. The nationality and its denomination of the captured banknote image are discriminated by the approximately nearest neighbor algorithm [ANN]. We proved the usefulness of the proposed algorithm by comparison the proposed with the SIFT matching algorithms of the execution speed and correct matching rate.