As the Covid 19 pandemic raged on, I was tasked – along with my work partner to create an elevated temperature detection machine with the help of the Intel Corporation – with whom I am a technical ambassador for.
The concept while not new – faced one glaring problem that the world was not aware of…. They did not work. Pointing a thermal camera at someones forehead is not a reliable way to measure for febrile illnesses. The cheap hardware and questionable software being produced at the time simply was not accurate enough, nor did it take measurements in a proper location – so we aimed to build a device that was grounded in science that simply worked as intended.
It is important to note here that I did not program this device – my work partner, who is quite possibly the smartest person I’ve ever worked with did the 1’s and 0’s. I was responsible for the product creation and fabrication of the device as we worked in lock step together.
Working with a medical cart manufacturer, I designed a portable machine, capable of being placed anywhere, that could fit through a doorway and measure subjects from 3 to 7 feet tall without needing to reposition the machine. To achieve this, I built a camera gantry on an automated linear slide rail that was controllable from an operator’s screen to match the height of the subject. From there, using an Intel RealSense depth sensing camera and a high end thermal camera we mapped the subjects face to find the ocular canthus – a small region of the eye near the tear duct. This is the most reliable place on the face to take a core body temperature reading.
To ensure our cameras were calibrated, we employed the use of a “black body” – a simple radiant heater element that remained at a constant temperature and served as a reference point for the thermal camera.
To operate – a subject needs to walk up to the machine, a screen facing the subject will instruct them to take a step forward or backwards and to remove glasses if needed.
Removing a mask was not needed as we could automatically detect this feature and did not focus attention on it. As soon as the subject is in the proper location, the machine would scan the canthus for 1-2 seconds and return the temperature to the operator side of the machine. If no fever was detected the subject was allowed to proceed – if an elevated temperature was detected, they would be taken to a secondary screening area designed by the machine’s owner. The machine was linked back to an analytical platform for remote statistics on the subjects.
We completed 10 fielded prototype of our machine – due to the pandemic, the construction of them all occurred in my garage and barn and were hand delivered to various schools and warehouse facilities. When we started the project, the market was flooded with extremely high priced “solutions” that promised the results we achieved without actually working as described. When the CDC released a report illustrating that the science behind these competitors did not reliably do anything – the market bottomed out. Machines ranging from $20-40K were now being sold for less than $5k, now you can find them for a couple hundred dollars.
When many failed, we decided to pivot – our solution is currently being explored for the next generation of smart buildings and could have a long future in that market segment. The big take away I learned from this project, aside from scaling a prototype into a production ready unit – is that you shouldn’t always question, what isn’t the competition doing – but are they doing what they say reliably. .