How can I measure the temperature of each pixel?

How can I measure the temperature for each pixel from a thermography image. I'm attaching an example of one of the images I took.

Can anyone help me?

Thanks.

[darova]

relationship color/temperature?

[Mokhtar]

Yes exactly. I need to link the color of each pixel to the temperature. BTW, I have a software that can tell the temperatrue of each point but I need a matlab code that can measure the area that has a certain temperature range based on the number of pixels. I hope that was clear. If not, please don't hesitate to ask.

Thanks.

[Rik]

There should be some metadata somewhere that links the colors to temperature. This is not visible in the image as an image. Can you attach the files you put in to that other software? We might be able to extract the needed metadata from that.

[Mokhtar]

I didn't quite understand your point but here is an example of an original picture taken using FLIR thermography camera.

Please let me know if you need more info and thank you so much for the prompt response.

[Rik]

Look at the extention of that file: fcf is not a normal image file. The temperature scale you can see in the program is probably encoded in that file. You can attach it as a file (you might need to zip it first).

Answer 1

[Image Analyst]

Of course it's better to use the actual image if you have one that has actual temperature data in it. But if you don't, like you have the cheapest FLIR one phone camera, then you can extract out the colorbar and use it to convert to temperature. See attached demo.

[Mokhtar]

Wow, this is just genius. Thank you so much man. However, how can I crop out the image and the color bar separately if they are in the same image? I mean, I can clearly see that you know the image row number and column number but I don't know the locations in mine. Would you please help me to figure out the location of the bar and the image? I'm attaching an example of an image I will be working on.

Side Note: the area of interest is the rectangular turbine blade. I know that the background is disturbing but this is how the FLIR camera takes the thermography images (always includes background).

Thanks again.

[Image Analyst]

The camera or software you're using will most likely put the color bar in the same location regardless what is happening in your scene. Therefore you can determine the location with any program (or imtool in MATLAB) and use those coordinates to crop out the colorbar.

Then everything to the left of the colorbar will be your image so crop it out

rgbImage = rgbImage(:, 1:leftColumnOfColorBar-1, :);

[Mokhtar]

Thank you so much Image Analyst, I managed to crop the image but I think the code couldn't read the colors of the area of interest. As you can see in the original image, there is a contrast between the cloro of the blade and the two areas of interest (kind of brighter). However, the code recognises everything as the same grey degree. This is what I got:

Any idea what is the reason behind this? and is there a way to fix it?

Thanks again.

Answer 2

[Guillaume]

You can't. Your thermography image is not the raw image from the sensor. It's a processed image that associate false colour to the raw signal.

Your software can convert the raw signal into temperature because:

• It knows the characteristics of the lens (in particular its absorption)
• It knows the exposure of your image
• It knows the response of the sensor vs wavelength
• probably some more dependencies that I've forgotten

Even if you had the raw image from the sensor, unless you know all of the above, you won't be able to convert to temperature. To get true temperature you also need to know the emissivity of the object you're imaging.

[Mokhtar]

But there is a video on youtube of a guy who created a Matlab code that can tell the temprature at every point for any image (even processed images).

Here is the link of the video:

https://www.youtube.com/watch?v=gxWvFtSbhAg

Thanks anyway for your answer and I still have hope to find a code that can do just that.

[Guillaume]

What that video shows is that he's got an image for which he already knows the temperature range. Once you know that the most intense pixel in the image is 106F and the least intense pixel is 81F, it's trivial to interpolate any other pixel intensity.

It's not a very robust approach though. First, it assumes that the response of the sensor is the same for all pixels (or that the spatial non-linearity has already been compensated for). Also, the temperature is displayed with 4 significants digits, which is completely absurd, particularly considering that the image is stored as jpeg which is a lossy format.

So, yes if you already know the temperature scale of your image, you could do something similar. I would use the true greyscale image instead of a false colour one though.

If you do not know the temperature scale of your image, then you need to know the calibration of your camera, all the details I've listed above.

[Mokhtar]

First things first, thanks for the prompt respons and the detailed explanation. As for my situation, I do have the temperature scale for each image. However, I'm not an expert in Matlab. Thus, I don't know how to write a code to do as you explained in your comment "interpolation". I'm attaching an image that shows the temperature scale obtained from the FLIR camera. Any help will be highly appriciated.

Thanks.

[Guillaume]

I'm going to assume that your image is saved in greyscale, not using false colour (otherwise your first step will be to convert the false colouring back to grey scale).

minintensity = 81;  %whatever is the background intensity
maxintensity = 106; 

rawimage = im2double(imread('someimage.png'));  %whatever image format you're using. DON'T use jpg. flr is not a supported image format
%depending on how the image format, you may need the following
rawimage = rawimage(:, :, 1);  %This assumes the image is greyscale

temperature = interp1([0 1], [minintensity maxintensity], rawimage);

[Mokhtar]

Hi,

I copied the code and changed the intensity values according to the image I am trying to test. I then ran the code but I couldn't find any outcome. It just keep typing the name of the image (test)> I am sorry again for my lack of knowledge in Matlab but can you tell me where can I locate the outcome(The number of pixels that have a specific range of temperature)?

Thanks.

[Guillaume]

temperature is the array (the same size as the image) that contains the temperature of each pixel. temperature(1, 1) is the temperature of the upper-left pixel. temperature(end, end) is the temperature of the bottom-right pixel.