Multiclass Classification of Rupiah Banknotes Based on Image Processing

Abstract

This research aims to classify the nominal value of Rupiah banknotes using image processing and classification methods. The research design was conducted by collecting a dataset of Rupiah banknotes consisting of 30 classes, each with 100 images. This research uses image preprocessing using Canny Segmentation to create object edges and clarify image details. The Hu Moments method, which describes pixel distribution and object shape, is used to extract special features from the image. Classification modeling is then performed using Decision Tree and Random Forest to classify banknotes based on the extracted characteristics. Model evaluation is performed by measuring accuracy, precision, recall, and f1socre performance and using cross-validation with k-fold=5. The results show that the Decision Tree method is able to classify Rupiah banknotes well. In the performance evaluation, the Decision Tree method achieved the highest accuracy of 86.83% and good precision, recall, and f1-score for several banknote classes. The Random Forest method also achieved good results, with the highest accuracy of 78.67%. The classification evaluation results show that the Decision Tree method is better than the Random forest in classifying Rupiah banknotes.