Paper Currency Detection System Based on Combined SURF and LBP Features

Abstract

Currency detection falls into the field of computer vision technology. Detection of currency is a helping hand for visually impaired people. Moreover it is also useful in surveillance system. In this paper we presented a paper currency detection system which can detect paper currency from image. Detection is based on training different currencies. At first we extracted SURF and LBP features of currencies respectively. Later we combined both features. Then trained them with SVM classifier. SVM as a classifier performs very well in training image datasets. After that applying sliding window technique on input image, we detected currency from an image. In this currency detection system we focused on only paper currencies of Bangladesh. Along with currency detection, this system shows number of currencies and also the total amount of currencies exists in an image. The proposed system is able to detect paper currencies in rotated positions also and it achieves an average accuracy of 92.6% in detection.