Thermal image processing

[AdamP]

Hello!


I'm working in image processing since a few years and currently I need to process thermal images. I'm new in thermal images and I would like to ask a few questions.

In the attached image the intensities in the center are missing. I heard that this comes from reflection. On this image I also saw small circular intensity difference, which appears on the same material. Is this also comes from reflection or this is normal behavior? Or this is because the camera needs to be thermally calibrated with a reference? Is there a way to decrease the error caused by reflection?

About geometrical calibration, I would like to ask that what kind of pattern is the best to detect corners precisely. I have read about using LEDs are not so precise. Another paper discussed about using a simple chessboard printed on a simple paper. Then they used extra light source to light the paper from different directions. I have found one more paper, where they used two different materials to detect a pattern. Can I extract precise corner features from a wood board and steel squares, like a two material chessboard? Can reflection decrease the precision of the calibration?

Finally I would like to ask that is thermal image processing different from simple image processing. If yes, could you please tell me where can I learn more about it. I have seen on the internet that there are thermography courses. Would it be useful to attend on a training to achieve better results in image processing?

Thank you for your help.

Adam Papp

 

[IRstuff]

No image?

What wavelength; LWIR or MWIR?

Missing center is often narcissus, wherein the optics of the camera reflect all or part of the sensor back onto itself. Generally, it shows up as a dark circular blob fuzz in the center of the image. This is not always normal thing, and the optics designer should have verified that there was no narcissus during design.

One-point or offset calibration can usually mask narcissus. You should be able to do a straight offset calibration by covering the lens with a pad of paper. If you then view a object that's different in temperature than the pad and see row or column variations, you need to do a two-point calibration, which does require viewing a reference source like a calibrated blackbody. The row or column differences are due to not correcting the gain.

These calibrations are often refered to as "NUC," non-uniformity correction. One-point corrects for offset; two-point corrects for both gain and offset. Depending on the camera, two-point corrections must be done if the scene temperature changes by large amounts, say, 10°C or 20°C. That's because a thermal detector's response curve is generally very nonlinear, and scene temperatures too far outside of the two-point calibration range will accentuate the non-uniformity of the gain across the sensor.

Depending on the accuracy of the distortion measurement desired, you may need to use a collimated point target and an accurate pan/tilt stage to mechanically measure the distortion. Ostensibly, one could use a rectilinear array of point sources, like an array of pinholes placed in front of a blackbody in a collimator to visualize and measure the distortion. Hypothetically, other targets as you describe could be used, but accurate measurements might be difficult or too manual. Depending on the optics, you might be able to use something like a shiny surface with black paint dots accurately placed on it with a heat lamp illumination. The difference in absorptivity will result in temperature differences that can be view by the thermal camera.

Thermal imagery is basically panchromatic, and is not much different than B/W photographs, other than the fact that thermal imagery will never be as sharp as visible B/W, due to larger diffraction blur, and the fact that you are imaging a emission rather than reflection, so "lighting" is generally irrelevant, unless you specifically want it so.

A calibration balckbody is one such case, where the surface of the blackbody is blackened to maximize its emissivity in the thermal bands of interest. This makes the emission more precisly a function of the temperature of its surface and mitigates reflections, which are usually only reduced, but never complete eliminated.

There are probably sufficient resources on the web to answer basic questions. Thermography courses are usually intended for users that are in the business of measuring precise temperature differences in images to determine locations of hotspots or thermal leakage through walls or insulation.

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize