[Rock-dev] 3d point uncertainty

[Alexander Duda]

On 11/07/2013 07:59 PM, Sylvain Joyeux wrote:
> On 11/07/2013 07:57 PM, Alexander Duda wrote:
>>  From my point of view it makes to add this kind of informations to the
>> point cloud. The only problem we might get is that the log files are no
>> longer compatible and must be updated.
>
> This last point is something we need sorted out. We are talking for a 
> really long time now about updating some core types, so we REALLY need 
> to make sure that the infrastructure to update log files is foolproof.
>
> Sylvain
Agreed, the only problem I see is who has time to do the tooling?

To not block the hole process we could also think of creating a new 
package for vision related types and deprecate the one we have in base 
(the tooling is already supporting this particular case). The idea here 
is that in this case we can more or less directly use opencv types 
without introducing a new dependency to base types. Furthermore, in the 
next couple of weeks I will also propose some new types to cover the 
following use cases which would go well with a new package and would 
also solve the problem Javier has with the point cloud data type:

* send image keypoints and descriptors to other components 
(cv::detail::ImageFeatures)

struct  CV_EXPORTS  ImageFeatures
{
     int  img_idx;
     Size  img_size;
     std::vector<KeyPoint>  keypoints;
     Mat  descriptors;
};

* send tracks (triangulated 3d points) to other components which can be 
seen as a point in a point cloud but have additional informations like:
   - an unique id,
   - covariance matrix
   - Jacobian matrix ?

* send matches to other components storing which keypoint of which image 
is associated to which track.
   - image id
   - feature id
   - tracks id
   - certainty

Comments are welcome.

Alex

-- 
Dipl.-Ing. Alexander Duda
Unterwasserrobotik

DFKI Bremen
Robotics Innovation Center
Robert-Hooke-Straße 5
28359 Bremen, Germany

Phone: +49 (0)421 178-456620
Fax:   +49 (0)421 178-454150
E-Mail: alexander.duda at dfki.de

Weitere Informationen: http://www.dfki.de/robotik
-----------------------------------------------------------------------
Deutsches Forschungszentrum fuer Kuenstliche Intelligenz GmbH
Firmensitz: Trippstadter Straße 122, D-67663 Kaiserslautern
Geschaeftsfuehrung: Prof. Dr. Dr. h.c. mult. Wolfgang Wahlster
(Vorsitzender) Dr. Walter Olthoff
Vorsitzender des Aufsichtsrats: Prof. Dr. h.c. Hans A. Aukes
Amtsgericht Kaiserslautern, HRB 2313
Sitz der Gesellschaft: Kaiserslautern (HRB 2313)
USt-Id.Nr.:    DE 148646973
Steuernummer:  19/673/0060/3