Correspondences problem

[Kota Toyosawa]

Dear Group,

I am working on 3D PTV with my experimental data and have recently completed the multi-plane calibration.
Currently, I'm adjusting the parameters to run the procedure, but I'm having difficulty making correspondences between four cameras. I have subtracted the background from the images, so they only show particles without any obstructions. I've also tried adjusting the "criteria for correspondences" settings, though the results have not been perfect.
I've tried all the approaches I could think of, but I still haven't been able to achieve good correspondences.

When I adjust the parameters, I can see some correspondences indicated by red, green, and yellow dots (representing matches with some of the cameras). However, the distribution of these colored dots are apparently not matched with 4 cameras.
I'm concerned I may have generated too many particles, but based on the "particle spacing displacement ratio" from some of the manuals on your website, it seems my setup should be feasible.

Here are the questions I'd like to ask:

・Is it possible to track all particles even with this dense particle distribution?
・Do you have any suggestions for addressing this correspondence issue?
・Could you provide any advice on adjusting the "criteria for correspondences" parameters? I have been modifying them based on observed results, but I'm still uncertain of the optimal values.

I attached the sample file of my work.

Best regards

[openptv]

attachment is missing 

[Alex Liberzon]

Please attach again :) 

On Friday, November 8, 2024 at 5:38:28 PM UTC+2 Kota Toyosawa wrote:

[Kota Toyosawa]

Sorry, I hope you can download from this link below:

https://drive.google.com/drive/folders/1OA6VypeFrPBo5wDZJ2xgdCHIkS-v-q9w?usp=drive_link

2024年11月9日土曜日 0:40:14 UTC+9 alex.l...@gmail.com:

[Alex Liberzon]

The calibration seems to be incorrect. I mainly wonder about the camera positions at -Z and about the fact that the glass is also on negative Z. 

Could you please draw a clear sketch of the experiment? size of the tank, wall thickness, camera positions? is it in water? or in air? 

On Friday, November 8, 2024 at 5:38:28 PM UTC+2 Kota Toyosawa wrote:

[Kota Toyosawa]

Thank you for your help.

I attached the positions of model and cameras. Actually two cameras are on positive Z and other two on negative Z, but in order to set the dots on right calibration points, I had to set them there...

If it's possible to change the aspect ratio of calibration points not the whole distance between dots, I might be able to set cameras at the right coordinates, but can we? 

As shown in this pdf, the size of the model is 500 mm(x), 380 mm(z) and 310 mm(y), wall thickness is 4 mm and it's in air.

2024年11月12日火曜日 6:06:42 UTC+9 alex.l...@gmail.com:

[Alex Liberzon]

Hi, 

We might want to get on zoom to discuss things in details and see maybe the photo of the setup. If it's in air - please change the n1=n3=1 and then the glass thickness is less relevant and one might even ignore it and use all n = 1 as a first step. 

Multiplane calibration is useful when the number of dots in each plane is by far larger than number of planes and it's impossible to do one 3D calibration. It seems from the first sight that this is not the case here and it's possible that the calibration fails with the low number of points in each plane. 

are there raw images of the calibration, before background subtraction, etc. ?

Alex

we can set up a zoom call, use alex dot liberzon at gmail dot com