Morphology-Based Banknote Fitness Determination

Abstract

Replacing unfit banknotes is an integral part of maintaining public confidence in currencies while maximizing banknote lifespan in public payment facilities. This paper presents a banknote fitness determination method which mainly focuses on soil and stain detection using images scanned with contact image sensors (CIS). Difference images between fit and unfit banknotes may be used to determine fitness. However, these images may contain erroneous edges since the CIS images usually have some alignment errors caused by scanning, printing, and cutting operations. To resolve this problem, we first categorized the soiling patterns into two types: large-and small-scale. Then we used two different morphological-based methods to eliminate the false edges by security features. After the soiling patterns were extracted, the fitness level was estimated by a maximum standard score. The proposed method showed promising performance when using the Euro and Russian banknote databases. INDEX TERMS Image processing, image classification, morphological operations, machine learning.