Abstract

The global health challenges of recent years have accelerated the demand for automated, contactless, and efficient health screening systems. Computer vision, combined with thermal imaging, has emerged as a promising technology for rapid temperature assessment in high-traffic public and industrial environments. This paper presents an intelligent thermal screening framework utilizing deep learning-based computer vision techniques for accurate detection, localization, and temperature estimation of individuals in real time. The system integrates infrared thermal cameras with RGB visual sensors to improve subject identification and reduce false positives caused by environmental heat sources. A custom convolutional neural network (CNN) is employed for face detection and segmentation, followed by thermal mapping to identify elevated temperature regions. Experimental evaluation on datasets collected from transportation hubs, hospitals, and workplaces demonstrates detection accuracy exceeding 96%, with latency under 200 ms per frame, enabling real-time operation. The proposed approach also incorporates adaptive calibration to compensate for ambient temperature fluctuations, enhancing robustness in diverse environments. By enabling non-intrusive, high-throughput health screening, the framework offers a scalable solution for epidemic prevention, workplace safety, and public health management in smart city ecosystems.