Bangladeshi Currency Identification and Fraudulence Detection Using Deep Learning and Feature Extraction

Abstract

There are persistent rumors about counterfeit money all across the world. There is a massive loop of producing counterfeit currency that is developing alongside technology. Counterfeit currency production has gotten easier and more advanced day by day, so the detection process has gotten more challenging. Utilizing a variety of user-friendly counterfeit detection tools or software is the only method to stop fraud. Many people still do not have access to these softwares or tools for detecting fakes. Therefore, some of these programs or tools are not accurate, dependable, and free. This paper describes a potential software solution for detecting fake Bangladeshi banknotes. The most important thing is that it is totally free of cost and usable by all average people. The primary purpose of this software is to identify different currencies and determine whether it is real or fake. In this paper, Convolutional Neural Network (CNN) and FLANN-based Matcher with the Scale-Invariant Feature Transform (SIFT) algorithms have been implemented in the deep learning process to recognize the currency and determine whether it is genuine or counterfeit, to make the entire process functional. 1, 2, 5, 10, 20, 50, 100, 200, 1000, and 5000 notes from Bangladesh have been used in this research. One of the most suitable methods to detect fake currency is color-changeable holographic thread, The Portrait of Bangabandhu Sheikh Mujibur Rahman and “Bangladesh Bank (in Bengali)” written on the level of intaglio ink, printed hidden below the middle of the note, “Amount in Bengali” at the right bottom of the note, Some parallel lines or objects of intaglio ink, and color changeable ink OVI and SPARK. Additionally, this procedure makes it possible to quickly identify fake money. To make the system quick and easy for everyone, the entire system has been developed into a mobile based and as well as a web-based application.