Diseño y construcción de un sistema clasificador de fresas basado en técnicas de visión artificial.

Abstract

This present research work aims to design and build a system strawberry classifier based on artificial vision techniques, which responds to the current problem where the producers do not have a system that supports the classification of the strawberry before final dispatch. For this purpose, the information was analyzed to determine product classification parameters, infrastructure design needs, control elements, and image processing. The implemented prototype consists of a hopper where the fruit is stored, a conveyor belt that allows its movement, and containers for final disposal. Elements were attached to the prototype, such as sensors infrared, servo motors, USB cameras, development boards, Arduino Uno, and Raspberry Pi 4B used for control and image processing. The algorithm was developed through the OPENCV environment based on the hue colored model, saturation, value (HSV) and area calculation; allowing the detection, recognition, state of maturity, and fruit size. The communication between sensors and actuators occurs in the stage of control, where the Arduino Uno is in charge of sending signals and executing orders from the Raspberry Pi 4B. To visualize the data obtained from the system, it is used remote access connection with a PC. With the tests carried out, it is determined that there are values that vary between 85% and 95% of certainty in the classification of the fruit and 90.5% of the effectiveness of the prototype, being acceptable percentages within its operation concerning similar investigations considering these percentages as a complete success. It is recommended to increase the test times to determine possible improvements to the prototype make a technical proposal to the government level to implement equivalent projects at the industrial level.