they have to have what we call a black body camera
They make a statement like this that shows they have no clue about what a blackbody is, and the audience is expected to believe they are credible?
The black body does not make the camera more accurate, it helps the camera better understand how the environment is effecting emissivity so it can compensate.
However, given how thermal cameras for temperature scanning are quickly proving themselves to be impractical and unreliable, I guess it makes sense to eliminate the cost and setup complexity of a blackbody.
Well I tested the solution. and they are correct they do not need the black body :)
because there is trick :)
First I was amazed , really good result. Than I took black body and stick the photo on it. I put the black body temperature on 38 degrees Celsius so I pretend fewer. First result was OK, second I had 37.9 than I had 37.8 and always it goes down. After 20 test I had temperature on their screen 36.5 .... even the fact the black body had temperature of 38.
So they are correct they do not need black body they are calibrating the camera with the people going around. There is kind of feature which is adjusting the temperature value for the operator.
so it is trick
If you put this device in the hospital where everyone has temperature you will have no alarm :)
They take average temperature of 10 people, than add "average offset" to some algorithm. After 1h we measured temperature of 35°C blackbody using their 160x120 camera and it showed 37.8°C, same thing with fake pictures.
I can confirm also this point. We tested several days because I was confused about this system and how it's working without blackbody. After that we tried to simulate fewer and system stops to work properly.
But if humidity in the room will rise, all readings will be much lower, so it will calibrate on those lower tempes. 20% humidity rise is like 0.7°-1°C difference for thier camera.
So you have to recalibrate it everytime humidity/ambient temperature changes. Also outside temperature will affect temperature in foreheads of people - so again camera will calibrate on this elevated/lovered temperature.
I'm interested in what their thoughts are on the new ISO standards and general industry experts that are weighing in and saying that a black body is needed to reach the accuracy levels specified on most cameras like this that support the use of a black body regulator.
We are doing the blackbody vs no blackbody tests. We already did one run and are going to do another run. To give a sense of our process, we do an initial run and then send it to the manufacturer (e.g., Dahua, Hikvision, etc.). We then wait for their feedback (maybe they say change the firmware, settings, positioning, etc.). Then we run it again with their feedback to minimize any dispute about the implementation of the test.
One scenario I can see is similar to people lining up at Home Depot to return generators after a hurricane. If the system is not catching any fevers, does it have value? You could argue the feel-good value, but you need a LOT of feels for $10K.
People may want to return these when they realize it requires ongoing calibration, someone to staff it, a process for dealing with flagged persons, and then in the end it never catches an actual infected person.
One scenario I can see is similar to people lining up at Home Depot to return generators after a hurricane.
I doubt that's going to be grounds to get an RMA approved.
I don't know how coronavirus is going to play out but it strikes me that people will want to use these systems for at least months, right? Though I am curious what happens in July when people are walking in off of the street when it's 90 degrees out (and how that will impact accuracy).
Though I am curious what happens in July when people are walking in off of the street when it's 90 degrees out (and how that will impact accuracy).
I'm actually extremely worried about this. I have alopecia and fair skin. 20 minutes outside unprotected mid-day = instant sunburn. I will be perma-banned from customer sites due to a never-ending "fever".
cameras are not living up to the value proposition pitched to the customers
In general, I have not seen that usually be allowed for RMAs (typically it's for things that are strictly broken, not, e.g., the WDR is not good enough for me, etc.). Also, even if it was, they would send a replacement, not a refund, right?
Any manufacturer is highly unlikely to issue an RMA for a $10K camera for a refund unless there is a massive justification like the threat of losing a much larger revenue stream or legal action if they don't do it. Or they have such high demand that they want it back to sell to someone else.
I suspect that most business owners don't really care how accurate these are. All they care about is if it will bring back employees and customers and mitigate liability if this technology becomes status quo.
I am wondering if Manufacterer #10 is saying that the trick is that they assume all the "hot" bodies are humans and then assume these temperatures as the normal human temperatures? ie use people as the blackbody reference.
If so, I can see why this can be a problem in many circumstances.
More generally, I am wondering: is the goal to identify people who have fevers and don't know it? Or catching people who have a fever and show up to work anyway?
I wanted to share some thoughts about the idea of calibrating "on the fly" with the last 10/20 people measured:
Shouldn't we break-up the calibration in 2 unknowns?
1 the delta due to the measurement environment
2 the offset between core temperature and skin temperature.
I don't see how one could "compute" an adjustment for environment factors without measurements of these environment factors (ambient temp, humidity, etc) and reliable data about the average skin temperature for that specific environment.
This is why I would argue in favor of a black body because it eliminates the first unknown which still leaves us a challenging second unknown to solve, which is how to get from skin temp to core temp. This is challenging but will be easier if we start from correct measurement data rather than algorithmic adjusted data.
Once we have good data, in situations where there is a lot of throughput one could argue that the offset could be automatically adjusted based on the expected average human temperature. However in situations where there is not enough throughput this can't work.