Non-contact infrared thermal imagers for mass fever screening (Part 3 of 3)

How accurate is Thermal Imaging?

For a thermal imaging instrument with NETD about 50 mK; a typical temperature measurement accuracy: 

±0.3℃ with blackbody    (a rough estimation by 0.05 times 6 to include 99% )
±1℃ without blackbody 

Absolute Accuracy is expensive to acquire and maintain

A blackbody radiator is basically a scientistic instrument that requires careful installation and calibration. The power cables and mounting racks have put too many restrictions on the scene and limited the applications. 

In a black-body installed system, the result of getting a false absolute accuracy is a catastrophe

With big data, deep learning and human sciences, the system can learn over time to become smarter and even more accurate in setting the new threshold, reducing False Negative primary and False Positives. 

It is not about accurate temperature reading, it is about how to screen out the fever population among the crowds.

Independent on the environmental change, an A.I. algorithm can always pick out the outlier and the febrile population without absolute temperature data, even though there will be a portion of false-positive (especially for female in ovulation period, there will be a ~avg 0.5-degree Celcius). 

Visitors who raised the alarm will be directed to another counter or quarantine section for further inspections. 

How large traffic can thermal imaging cope with?  

An affordable thermal imaging system can easily achieve over 250 people/mins still or moving, even though for a better flow management consideration, it is controlled to about 120 people per minute in real practice.

Who to contact?

For Chinese's material, installation in China : https://www.purpleriver.cn/
For HK and overseas : https://www.hk-rfid.com/contact-us-c5pg

No comments: