PIV uncertainty: Help and feedback wanted.

[William Thielicke]

Pretty often, you are asking for a method to quantify the PIV uncertainty. I implemented a method from 2013, and I need you help and feedback to finalize this. It is actually pretty simple, and my code is very straightforward I think. So anyone that had some math in school will be able to help. I (almost) didn't have math in school unfortunately, so I would be happy for support! If you want PIVlab to be improved, then just invest some of your time please.

Here is all the information with many images:

https://github.com/Shrediquette/PIVlab/issues/58#issuecomment-966591348

[Ademola Jayeola]

great progress! gotta go through it

[Thiago Borges]

I've been trying to test your code with my experimental data, but keep getting complex matrices for peakx and peaky after using "multispot_SUBPIXGAUSS" function. Would you know why that's happening?

Thanks!

[William Thielicke (Shrediquette)]

Hi, v2.60 changed quite a bit of the structure. But with version 2.59 the code works: Open piv_FFTmulti.m and look for the commented block of code that performs the PIV uncertainty calculation (I usually set a breakpoint before this block of code and then execute line after line to see what happens). The code would have to be slightly modified to work with 2.60, but this would eb a task that comes later...

[Tyler Mccormack]

How do you get the "intermediate results"  that you load in in your example?

[William Thielicke (Shrediquette)]

Hi,

I modified the code to make the disparity calculation again possible in 3.0 and later.

You would need to download this file and replace the existing one from your installation:

https://raw.githubusercontent.com/Shrediquette/PIVlab/main/piv_FFTmulti.m

Then, in line 237, set this variable to 1.

keep_all_images_in_memory=1;

An then just set a breakpoint in line 466 and execute the code that is commented out. It will give you a matrix with the disparity. But I think my code is still wrong as the disparity is way too high.

[Tyler Mccormack]

Thank you for the help. I see the large error, my disp_error is ~0.90. Just to clarify the units, this is ~90% error?

[William Thielicke (Shrediquette)]

This is in pixel units