Embedded Vision and Thermal Screening for Automated Door Lock Control

Abstract

This study presents the design, implementation, and evaluation of an automatic door locking system that integrates real-time face-mask detection with thermal screening to enhance public safety in indoor environments. The proposed system uses a single-board computer for image acquisition and processing, a convolutional neural network for mask detection, and an infrared thermal sensor for fever screening. Decision logic fuses visual and thermal inputs to automatically control door actuation: authorized, non-febrile and properly masked individuals cause the door to unlock, while individuals failing either check are denied access and logged. Experimental results from a laboratory prototype show mask-detection accuracy above 96% under varied lighting and pose conditions and fever-detection consistency within ±0.5°C of a reference thermometer. The integrated system demonstrates a practical, low-cost solution for access control during infectious-disease outbreaks. Unlike prior works that treat these functions independently, this study implements both modalities on a single low-cost embedded platform, ensuring real-time operation, privacy preservation, and scalability. The proposed system provides a practical, deployable, and contactless solution applicable to healthcare, education, transport, and corporate environments during and beyond pandemic conditions.