Visão computacional aplicada na classificação de grãos de milho

Abstract

This paper proposes to use computer vision to classify corn grains. Due to the subjectivity that the manual method imposes, using computer vision techniques and machine learning can help to reduce the errors caused by the subjectivity of the classifier. The proposed work presents a prototype acquisition with lighting control, the creation of an image bank, an exploratory analysis on the characteristics of corn grains and tests of 3 data sets in 3 known machine learning models, KNN, Random Forest and SVM, reaching an accuracy of 83% to classify three maize grain classes.