Amid the COVID-19 pandemic, body temperature monitoring is an effective measure to reduce the risk of COVID-19 spreading in the community. Two common ways to measure body temperature in public settings include non-contact infrared thermometers and thermal imaging systems. The use of non-contact infrared thermometers involves physical proximity, additional manpower and time, whereas the existing thermal imaging systems are either inaccurate or expensive. To address the above problems, an accurate and low-cost automatic fever screening system is developed to assist in preliminary health assessment. The system involves the integration of a RGB camera, a thermal camera, an embedded system board, and a cloud analytics platform. By utilizing temperature measurement and face detection, the system can extract forehead temperature and assess if the person being evaluated is wearing a mask or not. The system has conducted trial runs in various crowded settings including schools, restaurants, community centres, and commercial buildings. This study tries to lower the cost of developing an effective self checking station for fever screening by hardware trade-off and software optimization, to keep the cost within the market acceptance price. This solution is planned to be made widely accessible to serve our community during the COVID-19 period.