On Tue, Nov 30, 2010 at 6:56 AM, Thomas Moulard
<thomas.moulard@gmail.com> wrote:
> Hello everyone,
> I would like to start by saying that I'm sorry to ask a question that
> seems to be quite recurrent on this mailing-list
> but I didn't find the information I am looking for.
>
> I am trying to run stereo_image_proc on two PointGrey Flea2 firewire
> cameras connected on the same bus. There is no hardware (external)
> trigger.
> The camera documentation indicates it is possible to get synchronized
> images from theses cameras in this precise configuration.
>
> However, it is currently not the case as I am launching two different
> instances of camera1394_node to grab the images.
>
>
> My questions are:
>
> 1. I would like to know what are the plans to support stereo pairs of
> Firewire cameras?
> I think many people use similar configurations and it is really a
> shame that ROS does not work "out of the box" for this
> as for all the components I have tested so far.
>
> 2. How should this problem solved?
>
> I don't see any way to make stereo acquisition work with separate
> components doing the grabbing as this is done
> currently. Should a camera1394_stereo_node be written? Is it how this
> problem should be solved?

There are three possible solutions outlined in this enhancement ticket:

  https://code.ros.org/trac/ros-pkg/ticket/4217

Currently, stereo image processing requires identical time stamps for
the two stereo images. Running two separate mono versions of
camera1394_node will not ensure that.

The ticket suggests modifying image_pipeline to relax that requirement
slightly. I don't know what the status of that request is. No
committed fix is mentioned yet. I wonder whether it will be in
Diamondback. I don't see any image_pipeline or stereo_image_proc
reviews for that distribution yet. So, maybe it could still be done.

As ticket #4217 said, one good option would be to create a stereo
version of the driver, which could publish both images with the same
time stamp. This could be done in time for Diamondback, if there's a
consensus that it's needed. I don't mind doing the work, though I'd
need to borrow or acquire a suitable pair of stereo cameras for
testing.

That was not one of the future enhancements identified during the
C-turtle API review:

  http://www.ros.org/wiki/camera1394/Reviews/2010-05-27_API_Review#Conclusion

Some time soon, we need to hold a review for Diamondback enhancements
to camera1394. I'll announce that here when ready. Please participate
if you can.

Regards,
-- 
 joq