Revision history [back]

If you want to code something up, I'd recommend to write a node listening to rgbdslam's output topic, pass-through-filtering it to exclude floor and ceiling and otherwise just omit the z coordinate and render the 2D points to an image.

Would be an efficient way to go and should be coded in a day or two.

Things to keep in mind: rgbdslam has no notion of up and down, its Z axis is the up vector of the kinect in the first frame, because by default it takes the /openni_camera frame of the first image as the origin.

Regards, Felix