Automação do registro de frequência acadêmica e otimização energética nas salas de aula da Faculdade de Tecnologia

Abstract

The Faculty of Technology of the Federal University of Amazonas has several classrooms used in different shifts and with a high flow of people, which requires efficiet academic control and continuous use of electricity. In this context, manual academic attendance registration may result in inaccurate information, fraud, and lack of confirmation of the individual's presence. In addition, electrical equipment may remain on even when classrooms are unoccupied or outside scheduled class hours, causing energy waste. Given this scenario, this work presents the development of a low-cost intelligent system for access control, automated attendance registration, and reduction of energy waste in classrooms. The proposal uses facial recognition based on convolutional neural networks, image processing, and automation devices integrated with a Raspberry Pi. The system is implemented using Python, cameras, LCD displays, a database, a web interface, and energy control devices. Regarding the investigated techniques, different convolutional neural network architectures are evaluated to define the most suitable facial classifier, considering accuracy, loss, robustness to visual variations, and response time. The experiments include variations in image conditions, access control scenarios, attendance registration, and energy management. The results indicate efficiency and reliability in access control and attendance registration, in addition to estimated energy savings between 20% and 40% in the evaluated scenarios, demonstrating the technical feasibility of the proposed solution for modernizing academic control and promoting the rational use of electricity.