Accuracy Comparison of UV-filtered Indonesian Banknotes Denomination Recognition Systems

Abstract

As technology progresses, monetary transaction systems around the world are being continuously developed. Artificial intelligence as part of machine learning, especially, emerges as a new trend being used in transactions automation. This research is written with a purpose to propose a comprehensive comparison of accuracy, in recognizing denomination of authentic Indonesian Banknotes (Rupiah) using image processing methods and machine learning algorithms. This research is comparing accuracy between some classification systems designed using several known classifiers, using three kinds of image resolutions. From this research, KNN produced 100% accuracy, while the accuracy for SVM varied between 12.5 to 100% depending on the kernel used.