Fake Currency Detection Using Image Processing and Random Forest Algorithm

Abstract

It became relatively easy to make counterfeit currency as technology advanced, such as colour printing. Counterfeit notes are produced so precisely that it's impossible to tell the difference between a genuine and a fake. We can define counterfeit currency notes as copies of actual currency notes produced without the approval of the state or central authority and used to conduct criminal activities. A parallel economy is run with the help of this fake currency note. This parallel economy causes a country's economy to deteriorate, lowering the value of its currency in the global market. As a result, it became critical to build a comprehensive technique to detect counterfeit notes to decrease the flow of counterfeit notes into the market. The fake currency detection system for Indian cash using image processing and machine learning is discussed in this research study. In order to identify real currency notes based on the distinctive features seen on the Indian currency note of 100 rupees, this research uses image processing and the random forest approach. While omitting the texture and colour of the note, banknote identification is accomplished by comparing received banknotes to a database of previously learned images. On the sample size of 400 notes, we found accuracy of 84.25%, recall of 66.25%, and precision of 78.63%. The recommended solution outperforms convolutional neural networks in terms of accuracy and training speed while using a simpler methodology. To prevent inflation and economic loss, this approach can also be used to different types of currency notes.