Fiducial marker or camera pose estimation

2
votes

Let's say, we've got an Aruco marker and we want to estimate a camera pose. We can use estimatePoseSingleMarkers() function, which is based, as I understood, on solvePnP() function. So, help me, please, answer the next two questions:

  1. What are the reference points of Aruco (keypoints or features) used in PnP algorithm?
  2. How does this algorithm decide which points in the real scene and in the image should be matched? In other words, how the 3d-2d correspondences are done?

Thanks.

1
opencvpose-estimationaruco

1 Answers

0
votes

I am not too familiar wir Aruco, but I can answer the second question.

The algorithm doesn't do the job for you. You need to perform the matching and then feed them into the solvePnP function in a ordered form.

You might want to have a look at FLANN matcher