Ask Your Question

unknown points in point cloud captured by swissranger 4k

asked 2011-07-19 08:11:30 -0500

diannay gravatar image


Using the swissranger_camera node connecting through ethernet (running linux using ros- diamondback), I consistently see points along the edges of the field of vision of the swissranger when there are no obstacles in the way(see the following link to a picture of this) The points near the vertex of the cone by the sensor have an evident abundance of these points from an unknown source. The sensor has ~5m range and this weird noise we see is <~2m in front of the sensor where it is in a obstacle-free environment. Are there any suggestions to what is causing this problem and any solutions to this?

edit retag flag offensive close merge delete


thanks. changing the amp_threshold parameter helped and also the modulation_freq for a longer range reduced most of these unwanted points
diannay gravatar image diannay  ( 2011-07-21 09:49:47 -0500 )edit

2 Answers

Sort by » oldest newest most voted

answered 2011-07-20 17:22:34 -0500

arebgun gravatar image

What you see is normal and is an artifact of how Swissranger TOF camera operates. There are a few variants of SR4000 cameras rated for maximum distance of 5 and 10 meters, this range is called a "non-ambiguity range". Measurements are subject to a so called “ambiguity” or a “back-folding” phenomenon that is due to the periodicity of the signal that is used for the distance measurement. If there's an object outside of that range the distance measurement id ambiguous: it could be at x, or x+D, x+2D, etc. For example, for a 5m camera if an object is 7 m away, it will show up at 7-5=2m. See Section 8.3 in SR4000 manual for full explanation.

As Ivan said it is possible to filter those distance values based on confidence values as they will be pretty low.

Also, there's a few papers that attempt to deal with that problem:

Probabilistic Phase Unwrapping for Time-of-Flight Cameras

Multi-Frequency Phase Unwrapping for Time-Of-Flight Cameras

edit flag offensive delete link more

answered 2011-07-19 10:35:37 -0500

It's been a while since I looked at Swissranger data, but from what I remember, the output includes a confidence field. Have you tried thresholding based on that?

And to state the obvious, make sure there aren't any thumb-prints on the sensor's face.

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools


Asked: 2011-07-19 08:11:30 -0500

Seen: 382 times

Last updated: Jul 20 '11