Automatic Currency Note Recognition for Sri Lankan Currency Notes using ANNs - SLCRec

Abstract

Automatic currency note recognition has been researched in many countries around the world in the recent past with the advancement of new technologies. Even though many researches have tried to come up with an universal solution for this area of research it always tend to depend on the country's currency note characteristics and the success of the identi cation depends mainly on the extraction of features. Hence a generalized system is not applicable in every situations. Sri Lanka has not been involved in any kind of research or implementation of this kind yet and it has become a necessity for the country to start looking for an im- provement. Therefore the whole purpose of this research is to introduce a customized and well designed way of entering to this domain. The research is expected to have good results by conducting excessive experiments before coming to the conclusion and one of its primary goals are to overcome the ability to recognize badly damaged currency notes in circulation of Sri Lankan community. The proposed system comprises of image processing component and neural net- work component. Former component is responsible for extracting features and latter is responsible for taking correct decisions. Currency notes are not kept in one place and their condition deteriorate with age and hence a linear transformation function has been developed to meet these conditions and also provide robustness. Canny edge detector is used to as the feature extraction tool and sum of pixel values in rows taken as the input of the neural network. A three layer backpropagation neural network is trained for the classi cation with use of momentum ,threshold values and a sigmoid function. Coding is done using java programming language and all these components are combined to implement SLCRec system which performed well over what has been expected.