search

tomasz-lewicki / ai-thermometer

Fever screening with IR & RGB cameras and Deep CNNs
152 stars 41 forks source link

Calibrating Lepton 3.5 for Outdoor vs Indoor #23

Closed rgpettit closed 3 years ago

rgpettit commented 3 years ago

Has anyone run into calibration issues when using this outdoors - particularly in winter? The system works perfectly inside with ambient temperatures of 66-72F. However, trying to use it outdoors <45F produces really drastic readings on the facial detections - typically over 1000F!

I've tried leaving the system outside but am still having the same problems. I'm a track & field coach and am trying to set this up to screen athletes so I need it to work outside. Any suggestions would be greatly appreciated.

Thanks!

rgpettit commented 3 years ago

UPDATE - it's not the cold temperatures but rather sunlight that seem to be the problem. Works perfectly in my cold garage but as soon as it's in the sun, reading go crazy. It's not even facing the sun directly - just the presence of sunlight affects it.

nanogear commented 3 years ago

Probably this is your problem... emissivity:

https://www.flir.eu/discover/professional-tools/how-does-emissivity-affect-thermal-imaging/

tomasz-lewicki commented 3 years ago

Hi @rgpettit !

That's a really great application and I hope that I can be of help :slightly_smiling_face:

Lepton 3.5 is capable of sensing up to 150C in low-gain mode, and 450C in high-gain mode. It's therefore not capable of producing a (valid) reading reading of 537C (1000F), so I doubt that it's an environmental factor. The bug is likely caused by:

  1. Would you be able to share some visuals illustrating the problem? Could be a screenshot of the AI Themometer UI, or the output of this gstreamer command: gst-launch-1.0 v4l2src device=/dev/video1 ! video/x-raw,format=UYVY ! videoscale ! video/x-raw,width=800,height=600 ! videoconvert ! ximagesink
  2. Could you confirm which lepton version are you running (3.0, 3.5, radiometric/non-radiometric)?
rgpettit commented 3 years ago

Thanks so much for your help! I am using the radiometric Lepton 3.5 from DigiKey. Images attached from both the UI and gstreamer. "Normal" images were taken in front of a windowed door with full sun in the background. "Sun" images were taken standing in the same spot but with that door open. I get the same "normal" results anywhere indoors and the same "sun" results anywhere outside during the day.

Hope this makes sense. As you might notice from the UI image, I'm working with one-arm after injuring myself in T&F :-)

gst-Normal gst-Sun ui-Normal ui-Sun

nanogear commented 3 years ago

Hello,

Hope everybody is doing well. I tried to do the same today. Put the system in front of an open window(seeing the exterior light but not directly to the sun) and I am having the same results as @rgpettit

I am using the Flir Lepton 3.5(radiometric). Indoor works pretty good.

_Note: The gray/grey backgroud is because I changed the colormap to COLORMAPTWILIGHT but originally is black

test1 (sorry for the quality of the image, I was seeing over VNC at low quality)

nanogear commented 3 years ago

@rgpettit Hi, have you found a solution to the issue? Regards

rgpettit commented 3 years ago

Nothing yet, but I haven't had much time to look into it the last week or so. @tomek-l, have you discovered anything either?

tomasz-lewicki commented 3 years ago

Hi there!

I was able to reproduce the bug. It was due to uint16 type underflow during conversion from Kelvin to Celsius, and 328b29349f4f3e57665feb9bb87b527504821fd4 should fix this. I knew that Lepton 3.5 has (-10C to 400C) range, but I never really tested that -10C to 0C part of it in California :slightly_smiling_face: Checkout the newest master branch, and it should work!

rgpettit commented 3 years ago

Excellent! Great find - works perfectly for me now. Thanks!

nanogear commented 3 years ago

It worked for me too. Thank you very much!!!!