Combining pose estimates from multiple sources

asked 2018-06-12 22:22:36 -0600

malharjajoo gravatar image

updated 2018-06-12 22:38:17 -0600

Hi,

I am receiving 6 DOF pose estimates of an object from 3 cameras. I then choose a frame common to all cameras, ie, a world frame, and transform pose estimate from each camera to obtain pose estimate of object in world frame.

The pose estimates of the object in the world frame are - ( x, y , z, roll, pitch, yaw ) I have 3 measurements like this.

What is the best way to combine them ? I must emphasise I am not adding the Euler angles as I'm aware that they are not vectors, my purpose is to obtain a more robust reading since I have multiple cameras and hence I am taking the average for each of Roll, pitch and yaw.

Can I simply take average of the 3 values ? Can I average the roll, pitch and yaw values from each pose estimate ?

Thanks!

edit retag flag offensive close merge delete

Comments

Probably some keywords that could help:

  • quaternion interpolation / smoothing
  • SE3 interpolation / smoothing
  • SO3 interpolation / smoothing
  • [...]

Averaging Euler angles (roll, pitch, yaw) should be possible but may not be the best way to achieve this since this formalism should represent a sequence of rotation. Again, you should refer to the literature.

Eduardo gravatar imageEduardo ( 2018-06-14 16:12:20 -0600 )edit

Hi, thanks for point that out. Yes, I tried searching the literature but couldn't really find much on this. Also, why do you say that averaging the Euler coordinates will not represent a sequence of rotations ?

malharjajoo gravatar imagemalharjajoo ( 2018-06-15 01:59:52 -0600 )edit