forums.naturalpoint.com

Hi everyone,

I'd like to track the 2D locations and orientations of up to 6 objects moving around on the floor in a 2.5x3.5m area. Is there a simple way to take advantage of the 2D constraint (Z coordinate is constant) so that the cameras' FOV don't need to overlap?

I've set up our 3 camera (!) system so that the whole area is visible by at least one camera and there is some overlap in the centre. The calibration runs and places the cameras correctly but a marker will only be located if it's visible by at least 2 cameras.

Any advice or comments on this? We could buy 1 or 2 more cameras but beyond that it's not worth it for this particular experiment.

I could also use the raw marker coordinates from the cameras and simply project onto the ground plane but then I'd have to do my own calibration and I wouldn't have the benefits of the rigid body toolkit.

Thanks.

The software is currently designed to track a marker with several cameras, to develop a 3d capture volume.


You could look at adjusting the point cloud settings, and decreasing the min. number of rays, for a marker to be tracked with; but I'm not sure this is a perfect solution.

Hello:

I understand what you are trying to achieve, but I think it is best done by using the SDK to directly access the cameras and create your own projection.

I don't think you will have to spend much time on calibration. You could use 4 markers at the edges of the capture volume, and arrange your cameras to cover the area, then scale and apply the cameras coverage to match the 4 volume markers. This should give you a simple 2D "calibration". There is very little lens distorition in the cameras, so if you don't need exact locations, you can probably just ignore the lens calibration, and just do the transformation.

Does that seem like it would work?