[ros-users] image_proc doesn't work with camera1394

Jack O'Quin jack.oquin at gmail.com
Sun Mar 28 02:49:12 UTC 2010


On Sat, Mar 27, 2010 at 7:18 PM, Eric Perko <wisesage5001 at gmail.com> wrote:
>
> Thanks for the update. Could you post a launch file that works for
> you, grabbing just whatever the first camera plugged in is? I updated
> my version of camera1394 and verified that both /image_raw and
> /camera_info had nonzero timestamps, but image_proc topics still won't
> output anything using the launch file I specified in the original
> report. I'm going to try it on another system using a clean boxturtle
> install (I run latest on my laptop); if you have a system running
> latest instead of boxturtle, could you give it a shot and let me know
> if it still runs there?

I am using camera1394/launch/startcamera.launch. It loads the camera
data from a YAML file derived from the CAMERA environment variable.
Since it loads the YAML file from the camera1394 package directory,
you may want to modify the <rosparam file=.../> argument to point to
your own directory structure. To get the first camera plugged in,
leave the guid parameter commented out (like the test_camera.yaml in
the launch directory).

I don't see anything wrong with the launch file you posted.

As an experiment, you can probably use my startcamera.launch directly
with the (default) test_camera.yaml, but you'll need to run image_proc
in a separate terminal window. That's what I did today.

I've been doing all my testing with boxturtle, because I consider that
essential to getting good beta test reports from people like you. I'll
be trying it with latest soon, though.

>> On Sat, Mar 27, 2010 at 4:24 PM, Jack O'Quin <jack.oquin at gmail.com> wrote:
>>
>> I see that some versions of libdc1394 can return a timestamp for the
>> actual image capture. This is obviously better, I'll see if I can get
>> that to work reliably. If I can, I'll set that in the message header.

I just committed this change. It now uses the frame capture time (from
the kernel) as reported by libdc1394. That's the 1394 bus transfer
time, not the actual image capture time, but probably the best we can
do portably. I'm not even sure that will work on all systems.
-- 
 joq



More information about the ros-users mailing list