Compared RGB Methods Towards Efficient Money Detector for Blind People

Abstract

Limitations of profound visual impairment distinguishing each nominal number of banknotes are often used by people with bad intentions to take advantage of that basis, like money fraud. Due to this reason, the blind people need to be helped to recognize their surroundings by developing assistive technology that is advanced for them. This study aims to build an efficient design of a money detector by comparing three RGB methods: range breakdown, If-Then Rules, and decision tree to recognize the nominal of money. The sample used in this experiment is rupiah banknotes for the 2016 and 2022 issuances. The device is built with a TCS3200 colour sensor and designed in a real-time platform. It has been found that the highest average percentage accuracy was achieved by the breakdown range method with 100% (2016 sample) and 90% (2022 sample). This device also successfully produced a notification sound from a speaker that mentions the detected nominal value. This research could be used as a reference to improve assistive technology for blind people.