Welding arc's light saturate the image

Question

Today i come to you to ask you for your help because this is not my domain. I am a computer vision engineer and i want to monitor a WAAM (wire arc additive manufacturing) process using an OPTRIS PI 1M thermal camera.

The problem is when i try to record the welding process(arc is on) i get an image with white stripes/bands on the imageBad quality image . When i turn the arc off this is what i get Good quality image

When i asked the camera's manufacturer he told me that it's due to the recording frequency so i have to play with it until i find the best setup(he didn't suggest wheather the frequency should go up or down). I tried doing that but the image kept on getting worse.

In my humble opinion, I think it's due to the arc's light and i started looking for a filter to block the latter but unfortunately there are very few documents talking about thermal camera filters.

So here i am kindly asking you guys: have you ever encountered this problem ? How did you fix it ? What do you think the problem can be ? Any leads, suggestion, help ?

Thank you very much for your time.

Note: 1- The arc's is on(during welding process), its temperature value is always saturated (max value of 1800°). 2- I can eliminate this problem if i modify the reference bar (color scale of the image) by getting bigger the reference temperature (min temperature) (i make the reference temperature more than 950°C). Here is an image of the reference barReference bar

I tried modifying the recording frequency but it didn't work.

I modified the reference barreference bar and the arc's light almost disappear.

I'm currently looking for a filter to block the arc's light.

I'm open to all suggestions.

Answer 1

I don't think this is just "classical" saturation by overillumination. The stripes that you see come, I guess, from the way, the CMOS chip reads out the data. Somewhat from the bottom to the top in a striped fashion. Then during exposure, the other pixels see too much light. This could explain the repeated gradient.

Also I noticed, that you are on the very top end of the camera's temperature range: 1800°C should be the max according to the manufacturer, so maybe your process is too hot?

What I would try is:

1. Decrease the aquisition time of the camera further if this is possible
2. Increase the distance between object and camera (factor 2?), so it sees less light
3. Using a filter is a good idea, but I would assume that it changes the spectrum and hence you will get a different temperature reading.