Automated Fruit Grading System Using Image Fusion

This work proposed a recognition system capable of identifying an Indian fruit from among a set, established in a database, using computer vision techniques. The investigation made it possible to compare the image color models, together with the size and shape characteristics previously used by different researcher. For the class of fruits defined in this investigation, it was determined that the characteristics that best described them were the average values of the RGB channels and the length of the major and minor axes when the image fusion technique is used, a process that allowed obtaining results with an accuracy equal to 92% in the tests carried out, finding that not always selecting a greater number of variables to form the descriptor vector allows the classifiers to deliver a more accurate response. In this sense it is important to consider that among the study variables a low dependency or correlation value.