If anyone here has experience with infrared cameras maybe you could help me with a method to deal with the self‐calibration procedure that microbolometer‐based infrared camera performs periodically.
I'm trying to analyze the infrared signal from certain regions of interest in the frequency domain. Because the frequencies studied are in the very low range the recordings are of long duration like 15-20 min.
Because of the periodical shutter closures when calibration is performed there are spikes in the signal that will affect/make inaccurate the wavelet analysis performed on the signal.
Thank you.
It really depends on your intentions:
- You might filter out the wavelet analysis results where influenced by the calibration
- You might interpolate the values acquired during calibration - prior analyzing
- Some other methods might apply as well.
Anything that would be acceptable for you?
Dear Andrei,
The shutter for re-calibration necessary for most microbolometers. I do not recommend to turn it off because the pixel intensities will start drifting.
To calculate frequencies over multiple images, I recommend the following:
- First detect the calibration frames (I do that with a threshold on the average pixel value, as my "calibration images" are usually close to white)
- Replace all calibration images with zeros
- Calculate your frequencies as before
Nulling some of the Images is similar to nulling some samples of a time signal which you want to examine. This will lead to (hopefully very small) sidelobes in the spectrum, however your "main frequencies" are detectable as before.
If these sidelobes are problematic for your application, try interpolating the images as U. Dreher recommends. This will increase your mainlobe beamwidth (hence decrease frequency estimation performance), but has low sidelobes.
You should avoid simply deleting the calibration images, as this will affect the frequencies which you want to estimate.
I hope this helps,
Michael