Dynamic masking

[A Boggon]

Hi,

I'm trying to perform PIV analysis on the flows around a swimming object.

From what I have seen in papers on this subject (e.g. 'DYNAMIC MASKING TECHNIQUES FOR PARTICLE IMAGE VELOCIMETRY', Dantec Dynamics, 2017 ) the suggestion is to mask the object prior to performing the PIV analysis as the moving object can impact where the peak in the correlation is found.

I've had a look at the piv_FFTmulti script but the only part I could find the mask being applied was after the correlation is performed (what they call 'vector masking' in the above paper). Is this correct and the mask is being applied after the correlation or is it being taken into account before this point?

If I'm wrong and the mask is being accounted for during the correlation are there any suggestions about how to handle the fact that the object is moving? I would assume you should have masked the current frame and the next frame it is trying to find the peak correlation within but you can only hand a single mask per frame in PIVlab.

For context I'm using a slightly adapted version of the PIV_commandline script where I have a folder of frame specific masks that I can pass into the PIV settings (s{4, 2}) on each iteration of the main loop.

Apologies if this has been discussed before and I have missed it, if so please can someone point me in the direction of that discussion?

[William]

Here is the discussion that you could join:

https://groups.google.com/g/pivlab/c/6Ijm-McAihw/m/SY170t2mBAAJ

[Alex Liberzon]

in openpiv we implemented the path that was pointed in that discussion, see here. https://openpiv.readthedocs.io/en/latest/src/masking.html - it's not too difficult to convert to Matlab using chatgpt or claude.