<![CDATA[Advances in artificial intelligence and computer vision have enabled non-contact temperature monitoring through thermal imaging. However, facial temperature measurements may be affected by environmental conditions and individual characteristics, which can reduce prediction accuracy. This study develops a thermal camera-based facial temperature monitoring system using a Convolutional Neural Network (CNN). The system integrates an MLX90640 thermal camera, a webcam, a Raspberry Pi 4, an ESP32, a human-machine interface (HMI), and a servo motor to support automated monitoring. A total of 2,300 paired RGB and thermal images were collected and preprocessed into 128 × 128 grayscale images. The dataset was split into training and testing sets using an 80:20 ratio. The CNN model was trained for 60 epochs with a batch size of 16 and a learning rate of 0.001. Model performance was evaluated using Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). Evaluation across five subjects produced average MAE, MSE, and RMSE values of 0.128 °C, 0.032, and 0.168 °C, respectively. The actual and predicted average temperatures were 35.29 °C and 35.39 °C, respectively. Results show the CNN can estimate facial surface temperature with relatively low errors, supporting non-contact monitoring on low-cost embedded platforms]]>
Copyrights © 2026
