Control de nivel de líquido por medio de visión artificial utilizando un sistema embebido.

Abstract

Currently, the priority of many companies is based on the automation of processes, this involves increases in the level of production of either a product or service, it makes the operation line much more efficient in terms of machinery and raw material to be used, for this reason, the following project has as its primary objective the design and implementation of an intelligent system with artificial vision in the Raspberry Pi4 embedded card for the control of tank filling in a prototype in the laboratory of automation and control of the university of Pamplona (Figure 1), the intelligent system with artificial vision is supported with artificial intelligence from a neural network trained in a virtual machine configured with notebooks written in python language, with Google colab, the system was developed in three main phases, a main phase that performs the creation of a symmetric database fed by images, in the second phase is performed image preprocessing with the application of mathematical techniques, and image classification. In the database creation phase, images were first collected using the Raspberry Pi4 camera and each image was labeled with the type of liquid level, In the image processing phase, techniques were applied in order to highlight the areas of interest and create characteristic patterns which will be used in the image classification phase through the creation and training of a neural network model that was able to differentiate the liquid levels with high accuracy, also shows the detection of the percentage of level to control a ball valve automatically with a liquid level required by the user, In turn, the design and implementation of a graphical interface programmed in Python in synchrony with the trained neural network model and the detection of the percentage to control the level of filling tanks in the prototype from the embedded card raspberry pi4, finally a static web page programmed from scratch using Html, Css and JavaScript was developed, it contains the documentation that was necessary for the realization of this project, taking into account the need for people to access this information.