Hello-

I am using pcl to estimate the pose of object primitives (in this case, a
cylinder) from stereo data.

I searched ros.org for "kalman" and found the robot_pose_ekf package. I dug
around in the source for a bit, and it seems that it is all hard-coded to
only relate "base_footprint" to "odom_combined." It seems to me that with
some modifications, this could be used as a generic 6-dof pose filter.
Before I make a copy and go and change a bunch of things, am I right in
thinking this would work as a generic filter to smooth out the (rather
jumpy) estimates of the cylinder pose? The only incoming estimates would be
a visual odometry of sorts, that is, a 3D pose from the pcl cylinder
segmentation. Would this smooth that out even if it is the only sensor
source being used?

If I've overlooked some other handy package that will filter for me, that
would be good to know too...

Thanks!

Adam Leeper
Stanford University
aleeper@stanford.edu
719.358.3804