Re: [ros-users] ar_pose basics

Hey Prasad,

2010/11/18 Ivan Dryanovski <ivan.dryanovski@gmail.com>:
> Hi Prasad,
>
> Inline replies below:
>
>> 1. Does it stores a pattern and matches it when live camera is running?
>
> Yes.
>
>> 2. Can i work with only single camera OR a pair to get the marker? Does
>> camera_info topic requires stereo camera?
>
> You don't need a stereo camera setup. Since you know the size of the
> marker in advance, you can estimate it's 3D position with a single
> camera.
>
>> 3. Will i be able to work on ar_pose_ekf to estimate accurate position using
>> single camera? Since i have memory constraints so i am avoiding stereo
>> vision process. I have Hokuyo-URG-04 and Odometry sensors. I want to fuse it
>> with my existing sensors to get more precise localization inputs.
>
> The ar_pose_ekf was developed to work in conjunction with ar_pose. If
> you have more than one camera, the different estimates for the marker
> pose can be merged together using ar_pose_ekf. If you are only using a
> single camera, you probably don't need to use ar_pose_ekf (although it
> might help smooth out the result - I haven't tested it myself).

Like Ivan says, the ar_pose_ekf allows you to track a single marker
with multiple camera's. It's in fact a Kalman filter implementation to
get rid of the noise on the measurements, which allows you to use it
with a single camera as well to smoothen the estimates. The behaviour
depends on the configuration values which you put in (system model /
noise - measurement model / noise - etc). Let us know if you have any
more questions about it, we can give you some tips on how to set those
values.

Steven

>
>> 4. How it works in light different light intensities?
>
> It should handle reasonable variation in the light conditions. The
> best thing would be to just try it out.
>
>> I hope all my questions are relevant.
>
> Let me know if you need any further help!
>
> Ivan Dryanovski
> _______________________________________________
> ros-users mailing list
> ros-users@code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>