Estudio, diseño e implementación de un prototipo electrónico para la identificación de parámetros de calidad en naranja blanca, utilizando visión artificial.

Abstract

The present degree work consisted in the design and implementations of an electronic prototype for the identification of quality parameters for white oranges using artificial intelligence. For this, a controlled environment was designed coupled to a conveyor belt to be able to select the oranges by means of an algorithm using artificial vision in which OpenCV specialized library for artificial vision was used in a Raspberry pi four B embedded card, which was also used to control different actuators that make up the prototype. The process begins by releasing an orange through the conveyor belt and it is placed in such a way that it can capture a photo to later pre-process it, segment it and process it to acquire information on each of the oranges, having one object as a reference, the diameter is measured. Other parameter is the color of the orange for which, oranges that have a uniform color of ripeness and with a tolerance margin of 5% in the total surface are selected. With these parameters, it is decided if the orange is suitable or not for commercialization. Through the tests carried out for this prototype, it was obtained that it has a margin of error of 1.3% for the measurement of the diameter that corresponds to approximately 1.5 mm depending on each orange, for the selection of the color and the imperfections with an error of 3% in relation to a manual selection and with a power consumption of 23W. Based on the tests carried out, it is concluded that the algorithm with the method of inscribing the edge of orange in a circumference is the most appropriate for this application. It is advisable to incorporate an orange washing system prior to the prototype to improve the process and the effectiveness of the algorithm.