A Novel Classification Algorithm for New and Used Banknotes

With the development of global economy, the amount of banknotes is rapidly increasing. However, in the circulation process of banknotes, banknotes are readily torn or contaminated. The broken or contaminated banknotes should be retrieved by banks to keep the steady development of the economy. The banknotes sorting machine is a high-technology product with optical-mechanical-electrical integration, which utilize classification algorithms of new and used banknotes as their key algorithm. However, current existing banknotes classification methods suffer from time consuming and low accuracy. In this paper, through analyzing the feature of defects and contaminations in a gray-scale image, we propose an algorithm to classify new and old banknotes into five levels based on the statistical parameters of the feature points in a gray-scale image, gray level-gradient co-occurrence matrices and a multi-DAG-SVM classifier, which can be easily adopted by banknotes sorting applications. Experimental results show that the proposed method is a promising potential application with superior calculating speed and classification efficiency.