Many have tried to use this sensor for camera tracking and they’ve all come to the same conclusion. The positional data you get from it is far to inaccurate to be useful for camera tracking.

Is there any drifting or accumulated error using it for a while? Or vibrating when the sensor hold still?

My understanding is that it drifts a lot. 

Can the depth information of zed be made to support the tracking position of a person?

Bliss is good

Hi Solovai,

We don't have Stereolabs SDK integration, so Aximmetry likely can not receive its depth video/information. Also, we don't have any solution to track the position of a person from the depth information.
Though if you have a depth video, then you could probably put together a compound where you mask the depth video with the keyed billboard and get the average depth in that masked area. You can even use the Shaper module to shrink the billboard's mask, so there will be less chance of error. It is just a personal thought of mine, I don't know how well such a system could work.

Warmest regards,