Deformed banknote identification using pattern averaging and neural networks
Adnan Khashman, Boran Şekeroğlu, Kamil Dimililer · Computational intelligence · 2005
Neural networks combined with image processing can provide sufficient solutions to problems where automation and machine intelligence is required. An Intelligent Banknote Identification System that is able to recognise clean and deformed banknotes may be used to aid identification by machines or banknote counters. For such as system to be useful in real-life applications it must be also be fast, efficient and simple to use. This paper presents a fast intelligent banknote identification system that is able to recognise clean and deformed banknotes. Banknote image compression using Discrete Cosine Transform (DCT) and Biorthogonal Wavelet Transform (BWT) is used to simulate four levels of banknote deformation. The deformed banknotes will be used to test the trained neural network. A real-life application will be presented where this system is used to identify EURO banknotes and the new Turkish Lira (TL) banknotes. Experimental results suggest that the developed system perfoms well when using highly deformed banknote images as well as clean images; thus providing a fast, efficient system for recognizing banknotes.