Uma abordagem automática para leitura óptica de cartões: um estudo de caso

Abstract

This work proposes the development of an efficient software for reading multiple choice test response cards. In this sense, the developed system, using computer vision techniques, was able to recognize the candidates indicated alternatives as well as the candidate identification data. Regarding the methodology employed, image processing techniques were used, as well as the Python language, with the help of the OpenCV library. Thus, the present work was divided into five specific steps: image acquisition, segmentation, data reading / acquisition, exportation and results comparison. To validate the obtained results, the system was submitted to the reading of the response cards of the first stage of the 2018 entrance exam of the Federal University of Uberlândia. The results obtained in the tests were promising with high effectiveness rates and correct reading.