This paper presents a computer recognition system which is able to identify a Latin American tropical fruit, from a set previously defined in a database, using computer vision techniques. This study allows for comparison of the KNN and Bayesian classifiers and the RGB and HSV color models, along with the size and shape characteristics previously used by researchers in this area, in countries such as: Malaysia, Brazil and the United States. For the class of fruits defined in this research, the characteristics that best described them were the mean values of the RGB channels and the length of the major and minor axes, when the Bayesian classifier was used, a process that yielded results with an accuracy equal to 90% in the tests carried out. It was found that the selection of a larger number of variables, to form the classifier descriptor vector, does not always allow for the delivery of a more accurate response. In this sense, it is important to consider that there should be a low value of dependence or correlation among the variables studied. The project also resulted in the development of a calibrated video electronic scale prototype, capable of sorting fruit, device that aims to contribute to the solution of the problem of identification and classification of agricultural products in supermarkets. [ABSTRACT FROM AUTHOR]