Page 1 of 1

Setup and Calibration Questions

Posted: Thu Jan 11, 2024 1:05 pm
by loshea
Hello,

I have a few questions regarding camera setup and calibration.

1. On the website, it states that the optimal camera placement for a two-camera setup is 30-45 degrees. What would be the optimal setup for 3 cameras?

2. For the checkerboard placement during calibration, does the checkerboard need to be on the floor? We had the checkerboard on the floor and tried calibrating it several times but it said the board was not seen by one of three cameras. When the checkerboard was elevated by a step stool it was able to be seen by all three cameras. Also, in the picture for calibration (see attachment), the checkerboard is elevated too so I just wanted to know the explanation of where the checkerboard should be placed.

3. Where does the participant need to stand for calibration? Close to the camera? We are using OpenCap to study gait patterns so we have the participant start 10m away from the camera and walk towards them. Can calibration happen from that distance?

Re: Setup and Calibration Questions

Posted: Fri Jan 12, 2024 10:48 pm
by suhlrich
Please check out the information and videos here https://www.opencap.ai/best-practices

1) 2 cameras is fine for gait at ~30deg. If you want a 3rd, I'd put it just off of center.
2) the checkerboard needs to be perpendicular to the ground (see link above)
3) ~3m from camera is good

Re: Setup and Calibration Questions

Posted: Mon Jan 15, 2024 10:15 am
by loshea
Hi Scott,

Thank you for the reply. I have reviewed the best practices page. I understand that the checkerboard needs to be perpendicular to the ground but does it matter how far off the ground it needs to be? can it be elevated on a stool or table?

What are the benefits of using 3 cameras? What would be a situation in which more than two cameras are optimal?

Re: Setup and Calibration Questions

Posted: Tue Jan 16, 2024 10:41 pm
by suhlrich
The height of the checkerboard doesn't matter.

3 cameras adds robustness to occlusion, since every body part needs to be visible to at least 2 cameras at each timeframe; however, we've found that adding a 3rd camera doesn't necessarily add accuracy for activities where there is not much occlusion.