Closed pyrochlore closed 8 years ago
CalibrateEnsemble in the RoomAlive project renders more or less the same thing. If you try loading the example files in it, I think you will find the registration to be pretty good, so maybe you have a bug in your code? Alternatively, the transformation applied to generate mean.ply may be flawed, but I doubt that.
Thanks for the reply. Eventually I found the unit in transform matrix was in "meter" while the unit in point cloud ply file was in "millimeter". This is why the orientation of point clouds looks fine while translation didn't. The top view of the result of registration shown below Thanks!
I changed mean.ply to millimeters (develop branch, for now).
Hi,
I want to do registration to the point clouds output from RoomAliveToolkit. Since I didn't get my second Kinect v2 yet, I took the example provided in Readme for testing.
My process is simply read each mean.ply file below the camera folder, multiply the position of the points to the depthToColor matrix (found in ensemble.xml), then multiply all the position of points to the pose of camera (found in ensemble.xml)
The registered result I got is shown below points in red : left camera points in yellow: center camera points in blue: right camera
The top view is
From the top view, I can see gaps and shift between different sets of point cloud. Did I do anything wrong about the registration process?