Sistema especialista aplicado à inspeção da qualidade visual de grãos de feijão

Abstract

The visual properties of foods are an important factor for the definition of their quality and for determining their market price. For this reason, the inspection process for visual quality of food must be reliable, in order to attend to consumer expectations and to the standards set by the government. However, it is usual that this kind of inspection occurs manually and, if it happens, it can be slow and tedious; generate high operating costs; be subject to human failings, becoming the results standardization more difficult and making its automation desirable. In the case of the beans traded in Brazil, their quality is related to the amount of defects and mixtures that they can present in the evaluated samples. Their classification is made manually; it is based on rules and procedures established by the Ministry of Agriculture, Livestock and Food Supply (Ministério da Agricultura, Pecuária e Abastecimento - MAPA). In this research work, it was developed an Expert System ( Sistema Especialista - SE) applied to the visual quality inspection of grains of bean, in order to classify them according to the color of their skin/coverage. For this purpose, techniques of visual pattern recognition and artificial intelligence were used. Additionally, it was proposed a method for grain detection with the purpose of overcoming a limitation that is present in the recognition system of visual patterns available in the literature, destined to the analysis and classification of seeds and grains. This limitation keeps the grains spaced from each other, in or order to facilitate the process of segmentation and detection, which becomes its applicability in industrial processes more difficult. In the experiments done it was employed an apparatus composed of an automatic conveyor belt with a camera attached, in order to make the experimental environment as close as possible to those found in a production process. The average rates of success (98.62%), precision (98.50%) and accuracy (97.16%) obtained in the experiments indicate that the developed Expert System can be incorporated into industrial process of visual quality inspection of grains of bean.