Brain2: Machine Learning to Measure Banknote Fitness

A challenge for central banks is to decide whether used banknotes are still suitable for recirculation, or rather should be shredded and replaced by new ones. Obviously longer use of banknotes reduces the printing costs and environmental burden. Given the huge amounts of banknotes in circulation, determining the fitness of banknotes is not only of importance in cost control, but also poses a serious technical challenge in terms of processing speed and accuracy. With the current technology fit notes are often shredded along with unfit ones. In this paper DNB proposes a fitness detection method that may contribute to a reduction of the amount of unnecessary shreds. We suggest the use of the BRAIN2 fitness detector. BRAIN2 is the abbreviation of Banknote Recirculation Analysing & Instructing Neural Network fitness sensor. It is based on a machine learning method using a combination of intensity and contrast differences on colour images of the entire banknote. Studies done by DNB and the University of Amsterdam in 2010 have shown that this approach is successful in post-processing conditions. This study evaluates the performance of the BRAIN2 algorithm using two on-line systems: the automatic sorting machine CPS2000 and a double sided portable scanner with sheet feeder.