[ros-users] Simple navigation with USB-WebCAM
Bill Smart
wds at willowgarage.com
Mon Sep 27 22:20:35 UTC 2010
An additional problem with USB cameras is that, in addition to them not
being synchronized, USB is also a "best-effort" bus. It's not uncommon to
have droped frames, delayed frames, and other annoyances. I've used it in
the past with a static camera, and relatively slow-moving scenes with some
success, but there was a lot of pain involved filtering out the bad data
before we were able to do any actual computation on it. We were able to do
it because we were able to tolerate low frame rates and relatively large
position errors in the detected objects.
cheers
-- Bill
On Mon, Sep 27, 2010 at 2:01 PM, Kurt Konolige <konolige at willowgarage.com>wrote:
> Christian, good stereo algorithms get left/right image matches to
> subpixel resolution, so any relative motion of the cameras means bad
> distance readings. There are two effects:
>
> 1. If the cameras aren't synchronized, they capture images at
> different times. Even slowly-moving objects can cause several pixels
> of difference.
>
> 2. Even if the cameras are synchronized, rolling shutter introduces
> image warping from top to bottom. This is the same effect as with
> monocular cameras.
>
> Generally, moving the cameras is worse than having motion in the scene.
>
> Cheers --Kurt
>
> On Mon, Sep 27, 2010 at 1:49 PM, RobotNV <robotnv at gmail.com> wrote:
> > Adam,
> > Do you know of any data on how fast the movement of camera/objects need
> to
> > be compared to the speed of rolling shutter to influence results?
> > Aren't those shutters pretty fast? :)
> >
> > On Sep 27, 2010, at 12:45 PM, Adam Leeper <aleeper at stanford.edu> wrote:
> >
> > Hi Christian-
> > A cheap, easy stereo webcam is available from
> > Minoru. http://www.minoru3d.com/
> > It is about $60 on Amazon. It has a 6cm baseline, which makes it useful
> for
> > stereo range sensing in the range of about 30cm to a few meters.
> > Using the uvc_stereo package from umd-ros-pkg, it is straightforward to
> get
> > the stereo image feed and camera info published on ROS, and then all the
> > calibration and processing pipeline pieces work too.
> > The main disadvantage of these solutions, however, is that webcameras use
> a
> > rolling shutter, which means your stereo data is only reliable when the
> > camera and scene are static. Any motion, especially fast motion, will
> cause
> > skew in the image, which causes unreliable variations in the computed
> > distance to points. I'm not sure how this would affect your efforts to
> > navigate... I've been using stereo webcam stuff for mostly static
> analysis.
> >
> > --Adam
> >
> >
> >
> > Adam Leeper
> > Stanford University
> > aleeper at stanford.edu
> > 719.358.3804
> >
> >
> > On Mon, Sep 27, 2010 at 12:16 PM, Christian Mock
> > <christian.mock.chrimo at googlemail.com> wrote:
> >>
> >> Hello ROS-Community,
> >>
> >> I'm a total beginner using the amazing ros-environment...
> >> In the past, I've coded a lot of routines at simple ARM-based systems
> >> (Lego NXT and ARM9 Linux) using C and Lua.
> >>
> >> Last weekend I've made some tests with the brandnew rosnxt package and I
> >> had much fun with this Lego Gimick ... it works with ROS ! :-D
> >>
> >>
> >> Now, I'd like to start with a new platform (Ubuntu & ROS) and have some
> >> rookie questions to the community:
> >>
> >> For me it seems that ROS needs expensive rangefinders... True ???
> >> Or is it possible to use cheap stereo USB-CAMs for basic navigation ?
> >>
> >> Any suggestions how to use ROS without SICK or any other expensive
> lasers
> >> ?
> >>
> >>
> >> Thanks in advance
> >>
> >> Cheers
> >> Christian
> >>
> >>
> >> --
> >> Chrimo Private Google Mail Account
> >>
> >> _______________________________________________
> >> ros-users mailing list
> >> ros-users at code.ros.org
> >> https://code.ros.org/mailman/listinfo/ros-users
> >>
> >
> > _______________________________________________
> > ros-users mailing list
> > ros-users at code.ros.org
> > https://code.ros.org/mailman/listinfo/ros-users
> >
> > _______________________________________________
> > ros-users mailing list
> > ros-users at code.ros.org
> > https://code.ros.org/mailman/listinfo/ros-users
> >
> >
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ros.org/pipermail/ros-users/attachments/20100927/b71dac58/attachment-0003.html>
More information about the ros-users
mailing list