Classifying brazilian banknotes (R$) using smartphone image analysis

A new method to identify the authenticity of Brazilian banknotes (R$) is proposed using digital image analysis by a smartphone application, Photometrix®, associated with chemometric tools. Authentic banknotes (n = 4) obtained in local bank and falsified (n = 12) by the author himself by means of three printers (laser-Deskjet®, wax-Color Qube® and inkjet-Deskjet®) were used for the development of the methodology. An experimental planning was performed to evaluate the authenticity of the banknotes determining: the best region (holographic band, tactile region in the lower left and right upper number, effigy microimpressions, lower number left, animal figure, watermark and puzzle); Region of Interest (ROI) (16x16; 32x32 and 64x64 pixels), focal length (8, 12, 15 and 20 cm) and variation of the smartphone (Motorola G5 Plus, Samsung S7 and LG K10). The best regions chosen were effigy microimpressions and round the animal. Promising results were obtained from ROI 16x16 pixels with focal length fixed at 10 cm. Thus, the application presents promising features for analyzing the authenticity of banknotes by digital images making it fast, non-destructive, portable, with low cost, and can be accessed by all society.本研究提出一种基于数字图像分析的全新巴西雷亚尔（R$）纸币真伪鉴别方法，所用方案整合了智能手机应用Photometrix®与化学计量学工具。实验共使用从本地银行获取的真钞（n=4）以及作者本人通过三台打印机（激光台式打印机laser-Deskjet®、蜡基彩色打印机wax-Color Qube®与喷墨台式打印机inkjet-Deskjet®）制作的假钞（n=12）用于方法开发。本研究通过实验设计评估纸币真伪，考察的变量包括最优鉴别区域（全息带、左下与右上数字的触觉区域、人像微印、左下数字区域、动物图案、水印与拼图防伪区）、感兴趣区域（Region of Interest, ROI）尺寸（16×16、32×32与64×64像素）、拍摄焦距（8、12、15与20 cm）以及测试所用智能手机机型（Motorola G5 Plus、Samsung S7与LG K10）。最终筛选出的最优鉴别区域为人像微印与动物图案周边区域。当焦距固定为10 cm、采用16×16像素的感兴趣区域时，本方法取得了颇具潜力的实验结果。综上，该应用凭借快速无损、便携低成本且易于全民使用的特性，为基于数字图像的纸币真伪鉴别提供了极具应用前景的技术方案。