Overhauling de la estación de visión del MPS

Abstract

This titling work consisted of the restoration and implementation of the Vision station for the Modular Production System (MPS) laboratory, which said station was disabled and without any use, so it was necessary to reintegrate it into the system so that together with the other stations are used for learning and developing new research topics.

For the restoration of this station, an embedded system specifically a Raspberry Pi4 was used as the core, with a RAM memory capacity of 8GB, which makes its performance the most suitable for use in artificial vision control, and at the same time Possessing an Ethernet port, digital pins, and a Wi-Fi module facilitate communication with industrial protocols.

In order for the station to be able to recognize the machined parts from the laboratory, it was necessary to apply automatic learning supervised by the classification method, where 900 positive images of the parts and 500 negative images of the environment where they were going to be used as input data. These pieces were located, and it was required to use a Windows program capable of generating an xml file called Cascade-Trainer-Gui in which these positive and negative images were uploaded and thus the file to be used in Python was obtained.

In a matter of machine learning, the system was programmed in Python based on the OpenCV library, which interprets the xml file that contains the learning of the machined part, which, together with identification codes, makes its operation similar to facial recognition of smartphones, adding the possibility of locating and recognizing the pieces within a controlled environment.