2

I have a 3D object (a helmet) with a bunch a aruco markers on it. I'd like to treat these markers as a board. The markers are not co-planar with each other, but that is fine, per my understanding of aruco boards. The problem is, how do I initialize the board object coordinates (objPoints)?

It's not easy to take a ruler and measure their relative locations, since they do not all exist in the same plane. I could take a photo, detect markers, estimate the pose for each marker, and then figure out their relative locations from that. But I think doing this with a single photo wouldn't be very precise, nor would a single photo necessarily capture every marker.

Is there a common way to obtain objPoints from multiple photos for higher precision? Or is there any better way to do it?

Mark B.
  • 111
  • 1
  • 3
  • If someone could explain why this has been voted down, I would appreciate it! – Mark B. Aug 31 '17 at 02:02
  • Seems to be an [XY problem](http://xyproblem.info/) to me. Why do you want to do that ? Why not do a calibration with a standard planar grid to get the intrinsics and then do a [pose estimation](http://docs.opencv.org/3.3.0/dc/d2c/tutorial_real_time_pose.html) ? – Gabriel Devillers Aug 31 '17 at 07:55
  • I have already done camera calibration, so I have the camera intrinsics, and I've done pose estimation on individual markers. The problem with that is that pose estimation of individual markers hasn't been very reliable. Sometime the pose estimation is way off, and sometimes the marker isn't even visible. – Mark B. Aug 31 '17 at 16:47
  • In my testing with aruco boards, I get much better results. But putting an aruco board on a round helmet is proving a challenge. There's no space for a flat board. And while it appears aruco boards need not be flat, initializing the board parameters for such a board is a problem. So that's why I'm thinking I could take multiple photos, do the pose estimation for all markers seen, and then somehow process that data to get an accurate representation of the markers' relative locations to each other. (That last part is where I'm stuck!) – Mark B. Aug 31 '17 at 16:51

0 Answers0