Hi,

I am working on a project with a doctor from the local hostipal where I'm using ipisoft DMC to analyze the gait (only the lower limbs) of pacients.
The room where I did the filming is relatively small (we have to work with what we get..) and had to adapt the cameras setup to this issue.

The problem is that when I try to calibrate with the option auto detect initial cameras position, camera #2 and #3
appear way above their real height.

Real heights above ground :
camera #1 (right corner) - 1,28m
camera #2 (left corner) - 1,39m
camera #3 (middle) - 1,85m

Heights above ground obtained:
camera #1 (right corner) - 1,28m
camera #2 (left corner) - 2,39m
camera #3 (middle) - 2,45m

I have tried to manually adjust cameras 2 and 3 but to no avail. When i try to align the skin with the actor in one camera in the action video, it won't align correctly the other two.

Any help or tips would be appreciated!

Hello
Can you provide us your calibration video?

I've attached both the calib video (torrent file) and the calib project obtained from ipi Studio 2.
Thank you for the fast reply and support.

OK, started downloading. The download speed is very small, though, around 30KB/s :( Maybe, you have a small upload limit in your torrent client?

The internet speed is very low, unfortunately. The bandwidth allocation for the file is set on High and the upload speed is on unlimited but that don't seem to help..
I'm trying to add the file to dropbox, don't know if it will be better that way. I'll add the link here when completed.

EDITED
Here it is:
https://www.dropbox.com/s/9sbkexvw6lxwvy4/2013-07-04_16-18-54.iPiVideo

Thanks. The speed is pretty higher, around 1.5MB/s.

So, any luck so far?

Thanks for reminding, and sorry for delay. I almost forgot about this among other things.

First I want to note is a huge flicker on the image, caused either by the frequency of the light sources, or by electric power. This is not that for important for calibration, but surely will bring you much trouble for tracking of action videos. Try to switch the cameras to 50fps mode.
Second, the cameras seem to be too close to the actor, so in many areas he is not fully visible to all the cameras. It will greatly reduce the accuracy of tracking. You should think of finding the larger capture space.

I've processed the calibration video. At first, after the calibration has completed, there were indeed that wrong heights you mentioned. But there are two other things you should do manually to get the final scene:

• Mark the ground points
• Scale the scene based on either the real height of one of the cameras, or the distance between 2 cameras.

So, after I've marked the ground points and scaled the scene down to the height 1.28m for the camera #1, I've got 1.36m and 1.85m for other 2 cameras.
The project file is attached.

Just curious but how large is your room? I've used both PS3 Eye and Kinect, but these days I usually rely on the dual Kinect configuration because it requires less space.

Are you using a treadmill for the patients? I'm guessing you are since this will allow you to use a smaller space for walking motions, but I thought to ask just in case. BTW, if do you use a treadmill, it might help to have marks or some kind of flag to visually measure distance and speed.

Sorry, I know none of these ideas helps for solving your calibration problem but maybe some of it will be useful. Hope you get something working for you.

G.

_________________
Greenlaw
Artist/Partner - Little Green Dog | My Demo Reels (2013,) (2015,) (2017,) and (2019)

Watch a one minute excerpt on Vimeo now!

First of all, thanks for the support!
The project is way better than the ones I made. Still have some troubles with camera #2, where I can't seem to overlap the skin with the actor, altough the other two cameras are spot on. I'll now try to better ajust the orbits.

@Greenlaw The room is 6 meters x 5 meters, but there are obstacles like a table and some desks by the corner.
Unfortunatly I'm not using a treadmill, instead the patients are walking back and forth over 3 straight meters. Yeah, a treadmill would be appropriate, but wasn't possible.

If curious, a video of a pacient can be seen here: https://www.dropbox.com/s/etowztzebl1b1jf/2013-07-04_15-28-18.iPiVideo