Out-of-plane motion

classic Classic list List threaded Threaded
9 messages Options
Leb Leb
Reply | Threaded
Open this post in threaded view
|

Out-of-plane motion

Hi everyone,

First of all - great project!

I wanted to check regarding the units (I know it was mentioned somewhere on the blog that the standard units for motion is pixel/frame). This might be a silly question, but what are the units for the sheet thickness? I assume its mm, but I have no hands-on experience with macro-scale PIV, however, I seem to remember that in the PIV book by Adrian & Westreweel they mentioned order of mm for sheet thickness used in a "standard" PIV system).

While on the subject, could anyone tell me where does the "magic" integer 200 appear in lines 30 and 31 in the accuracy.m script? I am confused about what the z_move parameter actually means (some fraction of the sheet? millimeters? pixels?)

Thanks for your time,

Leb

William William
Reply | Threaded
Open this post in threaded view
|

Re: Out-of-plane motion

Hi,
Are you asking about the synthetic particle image generator in PIVlab? That is based on Raffel et al.'s PIV book - as far as I remember. So you should find an answer in there.

Please consider donating for the PIVlab project:
http://paypal.me/pools/c/81Y8wBVi4y

The developer needs to update to a recent Matlab release in order to keep the project running.

Leb Leb
Reply | Threaded
Open this post in threaded view
|

Re: Out-of-plane motion

Thanks for your reply. Yes, this is about synthetic image generator.

I do find their description of the image generation on page 165. However, they don't seem to actually do any measurement for OOPM. And I am not entirely sure how you implement your OOPM (i.e. purpose of the magic number: 200).

William William
Reply | Threaded
Open this post in threaded view
|

Re: Out-of-plane motion

Hi, I currently do not have computer access to check, but I think I did it like this:
The script in accuracy.m is copied from the main GUI.
The out of plane motion is given in percent to the script because of the GUI link.
100%/200% = 0.5

So with 100% out of plane motion, the particles in both images will come from two different random distributions (z0_pre and z1_pre)  --> there is no link between the two images anymore.

z0=z0_pre*(z_move/200+0.5)+z1_pre*(1-((z_move/200+0.5)));
z1=z1_pre*(z_move/200+0.5)+z0_pre*(1-((z_move/200+0.5)));

With 100% for z_move that gives:

z0=z0_pre*1+z1_pre*0;
z1=z1_pre*1+z0_pre*0;

Ok...?

Please consider donating for the PIVlab project:
http://paypal.me/pools/c/81Y8wBVi4y

The developer needs to update to a recent Matlab release in order to keep the project running.

Leb Leb
Reply | Threaded
Open this post in threaded view
|

Re: Out-of-plane motion

Ah, all clear. Thanks v. much!
Leb Leb
Reply | Threaded
Open this post in threaded view
|

Re: Out-of-plane motion

In reply to this post by William
I might be wrong, but I think because the distributions are random, there is a chance you will get an artificial match. In these images, you have z_move =  200% and the central bit does not change significantly.

A B
William William
Reply | Threaded
Open this post in threaded view
|

Re: Out-of-plane motion

Hi, that is true, but I can think of two reasons why this is not an issue:
1: PIV is matching particle patterns, and not individual particles
2: If you would have a real physical setup with 100 % out of plane movement, then the exactly same thing will happen: New particles enter the laser sheet and appear in positions that match previous particles that already left the sheet. The more particles you have, the more likely it is (imagine that every pixel is occupied by one particle).

I think (although I am not an expert in statistics) that this is normal behaviour of a statistical method like PIV.

How can you manage to get zero match between two random datasets? I currently can't think of a solution without putting artificial boundaries on the datasets (e.g. particles A must only be centered on even pixels and particles Bei must only be centered on odd pixels).

I think this is an interesting subject, and if you have further input, then I'd like to hear it.

Please consider donating for the PIVlab project:
http://paypal.me/pools/c/81Y8wBVi4y

The developer needs to update to a recent Matlab release in order to keep the project running.

Leb Leb
Reply | Threaded
Open this post in threaded view
|

Re: Out-of-plane motion

I agree with your points, and also agree that it is interesting. I might investigate this specific case further in my spare time.

Since you mentioned statistics, did you ever attempted to go through the "serious" statistical derivations for PIV? I tried multiple times, but it is way beyond my brain!
William William
Reply | Threaded
Open this post in threaded view
|

Re: Out-of-plane motion

I just did the "simple statistics" to validate PIVlab's performance. I analysed the effect of many different image parameters by simulating large amounts of data:

https://www.rug.nl/research/portal/files/14094707/Chapter_2.pdf

Please consider donating for the PIVlab project:
http://paypal.me/pools/c/81Y8wBVi4y

The developer needs to update to a recent Matlab release in order to keep the project running.