A Low-Cost IoT Node for Fever Detection using Artificial Intelligence

Abstract

Fever is the body's natural defense mechanism against infection. Baby's fever can signal a severe infection since they have more vulnerable bodies. Parents and caregivers are sometimes unaware of a rapid rise in a baby's temperature, which is a major concern among medical practitioners. Therefore, an automated real-time vision-based temperature monitoring for babies is proposed. The proposed system comprises 24x32 and 1080x30 thermal and RGB cameras, respectively, and a low-cost linux-based computing unit that utilizes a computer vision pipeline to detect the temperatures of both the baby and the parent. The temperatures are then compared to each other to reduce the impact of exterior environmental influences on the baby's reported temperature and improve accuracy. Results are then processed, analyzed, and transmitted to a system interface for visualization. The results obtained reveal that the system can successfully identify faces, distinguish between the parent and the baby's face, and alert the parent whenever the baby's temperature is above the average range with an accuracy of 83.33%. The system allows parents to seek medical attention if their baby's temperature continues to increase while asleep.

Keywords Computer vision; Monitoring; Temperature measurement; Thermal imaging