[ros-users] Localization from a camera (or two)?

[Elena Oleynikova]

Hi Max and Nathaniel,
In answer to the vslam question, no, it's not the same type as
slam_gmapping. slam_gmapping creates a 2D map, which you can then use amcl
and other 2D-based localization techniques on. vslam creates a 3D pointcloud
of the environment, and at the moment AFAIK, there is no good way to save
the pointcloud and then localize within it at a later time, as is possible
with the gmapping maps and amcl. There is a method for place recognition
within vslam_system that might be worth checking out if you want to continue
down this route.

An easier solution would be to do what turtlebot does for navigation: take
pointclouds from either the kinect or a stereo pair, convert them into a
laser scan, and use them with existing slam_gmapping and amcl applications.
Check out the pointcloud_to_laserscan package (
http://www.ros.org/wiki/pointcloud_to_laserscan ) and the
turtlebot_navigation package (
http://www.ros.org/wiki/turtlebot_navigation) to see how this is
implemented on the turtlebot. Though the implementation
is for a kinect, you should be able to do the same thing with a pointcloud
from a stereo pair.

Hope that was helpful,
Helen

On Sat, Jul 9, 2011 at 3:32 AM, Nathaniel Lewis
<linux.robotdude at gmail.com>wrote:

> I also have this question, for I have a Kinect and it would be cool to use
> some time of localization algorithms with it.
>
> Nathaniel
>
> On Fri, Jul 8, 2011 at 4:41 PM, Max B <skinkworks42 at gmail.com> wrote:
>
>> To All:
>>      I have a super-cheap robot that has one webcam, and may have two, set
>> up to be a stereo pair. Are there any projects working on getting
>> localization from an image source or a point cloud? Also, a related
>> question, does vslam create the same kind of maps that slam_gmapping
>> creates? When I watched the vslam video, it looks like it's a map of
>> tracking points, not a map of the environment.
>>
>> Thanks,
>>    Max
>>
>> _______________________________________________
>> 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/20110710/2de3eab0/attachment-0003.html>