Deer photos and Therm-App Pro data

[Ultrapurple]

There has been technical discussion on the Image Gallery II thread about an image I made of two deer.



The discussion starts here:

https://www.eevblog.com/forum/thermal-imaging/thermal-imaging-gallery-ii/msg3624069/#msg3624069

Several EEVblog members collaboratively reprocessed the image (starting from the raw data) and we ended up with a rather nice version:



Discussion then turned to the content of the data files that I posted and whether they represented an accurate view of the scene temperature. At that point we'd strayed a long way off the Thermal Imaging Gallery topic so I started this new thread.

I plan to make an image with ice-water and boiling water and image it with the same camera (Therm-App Pro) to provide more data on the relationship between the thermal values read from the sensor and those of the scene.

[JimM]

I plan to make an image with ice-water and boiling water and image it with the same camera (Therm-App Pro) to provide more data on the relationship between the thermal values read from the sensor and those of the scene.

Thanks Ultrapurple, that's a lot of work for you.

[IwuzBornanerd]

Edit to add: Those temperatures may not actually be the target temperatures. See this post from Stackoverflow:

https://stackoverflow.com/questions/56695676/how-to-open-a-txt-file-of-ir-temperatures-as-an-image-in-matlab-or-other-analysi and specifically the following:

"edit, since writing this I've found that the output .txt file of temperatures is actually sensor temperatures which need to be corrected for emissivity and background temperature to obtain the target temperatures. (One way to do this is to use the software which comes free with the camera to create new output .csv files of temperatures and use those instead)."

It is worth noting that, per one of the comments in that stackoverflow page, the camera used is a Therm-App TH.  Somewhere I got the idea that the model TH is the only one for which the company's own app provides temperature values.  I stopped concerning myself with what the Thermviewer file numbers are because I figured if the manufacturer itself does not provide temperature values for a particular model of its cameras the author of Thermviewer is not likely to get them right.  While the particular lens that is attached at the time is certainly a factor that needs to be included, it seems that one would have to also know how the sensor is "tuned" in order to obtain temperature values from the raw data.

[Bill W]

Assuming it really is raw sensor data (ie 14 bit ADC readings) then rather than knowing how the sensor is tuned you simply ...    ....need the raw data that would be used for a standard  manufacturer calibration.

The temperatures do not matter much, but using the usual quoted ones you need 3 frames:
Data (20°C flat scene)
Data (35°C flat scene)
Data (real scene)
Then simply crunch the numbers for each pixel for which ImageJ 'image calculations' or even a spreadsheet will do.

Typical ImageJ screens attached, from an ASi sensor, for the first step of (scene - 20°C flat) which is a remarkably credible image given there is no gain correction in there at all.

You then convert to temperature by multiplying each pixel value by its' corresponding [35 flat - 20 flat]/15 and add 20.

That works OK for around ambient, for further away you need to realise that the calibration is in terms of 'energy of a temperature of' not actual °C, so you need to correct for Planck curve effects or get very silly values once going much over 100°C.

Bill

[JimM]

It is worth noting that, per one of the comments in that stackoverflow page, the camera used is a Therm-App TH.

Good find. That may explain another point. In the Matlab code in the stackoverflow page, the image had to be rotated 90 deg to get the correct orientation. However for Ultrapurple's Therm-App Pro image you needed to mirror the image to get the correct orientation. I guess different software for the two cameras.

[JimM]

I plan to make an image with ice-water and boiling water and image it with the same camera (Therm-App Pro) to provide more data on the relationship between the thermal values read from the sensor and those of the scene.

I don't know if you have embarked on this project yet, but I don't think this will resolve the issue. Undoubtedly, the temp.txt file will show something near 0 deg C for the ice-water and something near 100 deg C for the boiling water. However, I think there are too many unknowns to determine how to modify the temps to get to the correct temps. For that matter how do you know that the ice-water is exactly at 0 deg C, etc. I was hoping someone had documentation that explained the temp.txt file rather than having to do reverse engineering of the file. (Not that there's anything wrong with reverse engineering).

[IwuzBornanerd]

Assuming it really is raw sensor data (ie 14 bit ADC readings) then rather than knowing how the sensor is tuned you simply ... 

Bill

Thanks Bill; it was the characteristics of the curve for the particular sensor that I was thinking of regarding how the sensor is tuned.  I also was thinking that there was no flat-field data, but I now recall that Ultrapurple has in the past said there are quite a few calibration files for the camera.  And I also thought there was no SDK for Therm-App, but according to the description of Thermviewer it was based on an SDK, so I suppose the SDK "knows" how to use the cal data, and therefore Thermviewer could be generating legitimate numbers, but then the question is whether they incorporate the emissivity, etc. mentioned on the stackoverflow page.  Someone who has purchased Thermviewer should make a "support" request...

[IwuzBornanerd]

I was hoping someone had documentation that explained the temp.txt file rather than having to do reverse engineering of the file. (Not that there's anything wrong with reverse engineering).

Does anyone on this forum have the ThermApp SDK such that they could see if the SDK doc's provide such information, or has anyone asked the Thermviewer author for that info.?

As for sample images, including something from the freezer would be useful.

[therwp]

If it would help, I am happy to contribute a second set of maybe-calibration images from my ThermApp Pro.

[Ultrapurple]

If it would help, I am happy to contribute a second set of maybe-calibration images from my ThermApp Pro.

I would appreciate that. The files should be in your ThermApp (or ThermViewer) directory, in a sub-directory named the same as your camera's serial number. Zipping them together reduces the file size.

If you don't want to post the files publicly, please drop me a PM and I'll give you my email address.