Sistem Deteksi Nominal Mata Uang Rupiah Menggunakan Metode Haar Cascades Classifier Untuk Penyandang Tunanetra

Abstract

This study aims to detect denominations of the Indonesian currency from facial images on banknotes using the Haar Cascade Classifier Method. The output of this research is sound to help blind people with physical limitations to differentiate the denomination of rupiah currency. The method used in the process of identifying the denomination value is the Haar Cascades Classifier. The Haar cascade classifier consists of black and white boxes to handle gridded images, where there are multiple pixels in a frame. Each box will produce a different value to show the light and dark values ​​as the basis for image processing. The money data used in this study is divided into 3 conditions, namely (1) New Conditions, (2) Semi-New Conditions, (3) Not New Conditions. The data uses images of Rp. 100,000, Rp. 50,000, Rp. 20,000, Rp. 10,000, Rp. 5,000, Rp. 2,000, and Rp. 1,000 banknotes. The data is then tested under different lighting conditions and distances, namely low light and sufficient light conditions with distances of 10cm, 15cm and 20 cm. The test results show that the optimal distance for the currency denomination detection system is at a distance of 10 cm with an average accuracy of 100% for sufficient light conditions and 95.3% for low light conditions.