Device that allows the identification by type and size of the naranjilla and the tree tomato by artificial vision to improve production in SMEs

Abstract

The present work consisted of the design and implementation of a prototype that allows the identification by type and size of the orange tree and the tree tomato, to achieve this, a computer vision system based on neural networks was used. The prototype has three systems: the image acquisition system consisting of a webcam and lighting for data capture; the artificial vision system uses the image processing method to perform the classification by size and neural networks for the classification by type, finally the mechanical system is made up of a storage system and a conveyor belt. For the training of the Artificial Neural Network, the convolutional algorithm was used that through the acquisition of images trains the identifier, three categories were established: tree tomato, naranjilla and unknown object, each one has a database. The programming algorithms were developed in Raspbian, with Python programming language using OpenCV and TensorFlow libraries. The graphical interface was developed using the Tkinter library, allowing the user to control six buttons: start camera, take photo, upload photo, fruit type, fruit size, close program. Based on the tests conducted, it is established that the system works optimally with 95% when detecting the type of fruit, 93% to identify the size, the response time ranges between 0.3 and 1.3 seconds. It is concluded that the prototype achieves the identification of the fruits by their type and size with a minimum margin of error.