Seleção e classificação inteligente de mangas por análise de imagens.

Abstract

Brazil is today one of the major world producers of fruit, where the mango is presented as an important component of its exports. To remain at the forefront of this market, efforts are multiplied in order to provide the consumer high quality fruits, where advances in information technology allow the development of automation systems for tasks of major import ance involving cognitive aspects, such as selection and sorting of fruit, which ensures a higher yield. Given this premise, this paper presents the development of a non -destructive automation system for classification and selection of Tommy Atkins mango, employing a computer vision system associated with a machine learning system that can specify and use any trade standard for review by color and shape. The estimation of parameters such as mass, volume, position on the stem and other relevant information are also implemented. This system was developed glimpsing the possibility of being embedded into a low cost classification conveyor belt, for the benefit of small farmers, where the developed interface allows operation by operators with little qualification.