Otimização do cadastro de notas fiscais por meio de aprendizado de máquina e reconhecimento óptico de caracteres

Abstract

This project proposes the development of an automation system for registering supermarket invoices donated to Non-Governmental Organizations in the state of Paraná. With the aim of simplifying and speeding up the process of inserting invoice access keys, reducing the need for manual entry of 44-digit numbers into online systems such as Nota Paraná, the system uses computer vision and optical character recognition methods. The steps include capturing invoice images using a mobile device, pre-processing to improve quality and readability, applying a Machine Learning model to detect the region of interest containing the access key, and extracting the text . The extracted access keys are stored in a text file for later use in registering invoices. The expected results include greater agility and precision in obtaining access keys, contributing to the improvement of control and management processes for donations of invoices by organizations. It is concluded that the proposed system offers an efficient and scalable solution for dealing with large volumes of donated invoices, significantly reducing the time and effort required for manual registration.