Reconhecimento de caracteres em imagens de anúncios de produtos

Abstract

Comparing product prices when shopping is an important activity that is part of our daily lives and that in recent years has been facilitated with the emergence of software that seeks to bring its users up-to-date data on offers and promotions at supermarkets. One of the softwares created for this purpose is LudiiPrice, an application for mobile devices that has as main function to perform the search, monitoring and comparison of prices of products from several establishments having, as main source for its database, a collaboration with its users who register recent purchases invoices through QR Code reading. To expand its database, LudiiPrice seeks to increase its source by searching for offers and ads in supermarket flyers on web sites and performing the extraction of the data present in these flyers. To accomplish this goal, this work aims to analyze techniques and strategies for detection, recognition and extraction of characters in cutouts of product prices in supermarket flyers offers, seeking to improve the quality of the images of the cutouts to obtain the highest possible accuracy and reliability in data extraction.