The Effects of Segmentation Techniques in Digital Image Based Identification of Ethiopian Paper Currency

Abstract

Paper and coin are the two most common currencies in all over the world. In Ethiopia also paper and coin currency are used for medium of exchange. This paper presents the comparative study of segmentation techniques towards Ethiopian paper currency classification. Otsu, FCM and K-means segmentation techniques are considered for this study and BPNN is used for classification of currencies. For the classification, images are collected from commercial bank of Ethiopia and Dashen Bank; for our data set, a total of 500 images samples were collected. From these images, 91.2% accuracy is achieved when Otsu segmentation is used on BPNN with TANH learning function.