Page 1 of 1

3D Digital Image Correlation (DIC) using StereoPi

Posted: Tue Jun 23, 2020 10:33 am
by KevinMoerman
I am working on a 3D digital image correlation (DIC) set-up which uses the StereoPi system. DIC is a technique to analyze images containing contrast such as speckles to compute 3D locations (shape) and if tracked over time offers the ability to measure 3D displacement. Gradients of the full field 3D displacement data then offer strain and other derived deformation data which is relevant to the biomechanical research that I do.

While at the MIT Media Lab Dr. Dana Solav and our team built a large 360 degrees 3D DIC system featuring a circular array of Raspberry Pi's and cameras. This was for 3D shape and 3D displacement (and deformation) imaging e.g. to scan lower limbs to build custom prosthetic devices. (accidentally it also enables matrix style "bullet time imaging").




My plan here is to create a small and low cost version of the above featuring only 2 cameras. This would be useful for imaging deformation during mechanical testing procedures (e.g. tensile uniaxial testing). Besides the simplicity of the StereoPi it also seems to offer the closest thing to true simultaneous imaging, which is really difficult to get right when aiming to trigger a large array of separate Raspberry Pi systems. Therefore I'll also aim to push the imaging speed to image more rapid deformations that were previously available.

What I've done so far:
1) Created a system to mount the cameras separately (not the regular parallel stereo configuration)
2) Added time-lapse photography acquisition button (and associated settings in the configuration file) to the index.php network interface

Next steps:
1) Perform DIC calibration steps, using dotted calibration object and checkerboard pattern
2) Apply a speckled pattern to a test objecct
3) Acquire a time-lapse series of images of the speckled object while it is deforming
4) Optimize resolution/imaging speed to push the maximum speed
5) Deployment of the technique to study soft tissue deformations e.g. the foot (to understand biomechanics of diabetic foot ulcers) and also the breast during clinical mammography procedures.
6) Publish the set-up in a journal like the Journal of Open Hardware (

I'll posts updates here (and on Twitter @KMMoerman as I make progress and will shares codes on GitHub.

Re: 3D Digital Image Correlation (DIC) using StereoPi

Posted: Tue Jun 30, 2020 1:15 pm
by Realizator
Hi Kevin,
I read your post several times, and got a lot of emotions! :-) Actually you are working on a very interesting project!
Your idea of "imaging deformation during mechanical testing procedures" reminds me of the hi-end cameras equipment, like Aramis. I have plans to write a review article about it to show a possible implementation of the stereoscopic and multi-camera vision. Your project goes right in this direction and intended to solve difficult real-life tasks.

Can you please give us some more details on your approach?
- Which cameras do you use in your project? (V1/V2/HQ, FOV)
- You mentioned "dotted calibration object and checkerboard pattern". If calibration performed in Matlab?
- "Optimize resolution/imaging speed to push the maximum speed" - do you need this for the real-time calculations, or to get more details on object transformation for the future analysis?

If you will have some StereoPi specific questions - please let us know. We'll do our best to help you!

Re: 3D Digital Image Correlation (DIC) using StereoPi

Posted: Tue Aug 25, 2020 8:30 pm
by KevinMoerman
Thanks for your interest in this project. Any updates on your article? I've just started to pick this project up again (my StereoPi was in use at a clinic for some experiments).

Here are some answers to your questions:
* For now I have two V1 cameras to use with the StereoPi (I bought the starter kit). The set-up we used at MIT and in those papers I linked to uses 16 Raspberry Pi 2 systems and 16 V2 cameras (but synchronizing them was challenging and not very precise at times, which is what got me interested in the StereoPi approach!). I also have two V2 cameras here and may buy the HQ version.
* Yes currently the toolbox developed by my colleague is a MATLAB toolbox. However I'm applying for a grant to develop a Python/Julia implementation.
* I do not need real time analysis of the deformation (post-processing of images is fine for now) but it would be great to image fast enough to study reasonably "dynamics" deformations. I study tissue biomechanics for which pushing the imaging to about 10 frames per second would be great. Writing the files to the flash card is probably the limiting factor (?). I've bought the fastest flash card I could afford so will test how far I can push it soon.

Thanks for your offer to help with questions. Let me know also if you need more information regarding that article you were/are putting together.