VTK/Geovis vision toolkit

Requirements

There are two parts to the project:

• Video / related meta data rendering and interactions.
• Displaying video / meta data in geospatial context.

As per our discussion we have gathered following requirements for the project.

• Easy to put together.
• Provides 2D and 3D mode where in 2D it will allow quick validation, manipulation abilities.
• Provides the ability to visualize, video, track, and associated ranking.
• Ability to provide input as sequence of images or a video file.
• Provides the ability to manipulate track data. User should be able to insert a new one, modify the existing one.
• Provides the ability to select a region with pixel level accuracy and draw a track on top of it.
• Provides the ability to play many video files (small or large video) files.
• Provides a layout for the videos in space so that each can be seen clearly for a view point.
• Allows the videos to play individually or globally.1
• Provides the ability to toggle different layers of information.
• Provides geo spatial context.

Design

• vtkVideoElementRepresentation would contain all the information layers and make sure that they are in sink.

Also this is the representation thats responsible for consuming the data for each video.

• vtkVisStrategyRepresentation

<source lang="cpp">

// Interface for this class is not as clear as it should be. Basically it transforms all layers // associated with any given vtkVideoElementRepresentation and represents this transformation via // some geometry. class vtkVisStrategyRepresentation : public vtkRenderedRepresentation { public:

 // Return transformation matrix that can be used to offset any other representation. 
 vtkMatrix4x4 GetMatrix(); 

 virtual void PrepareForRendering(vtkRenderView* view); 

};

class vtkVideoElementRepresentation : public vtkRenderedRepresentation { public:

 void AddTrackRepresentation(vtkTrackRepresentation* tr); 
 void AddVideoRepresentation(vtkVideoRepresentation* vr); 
 void SetVisStrategyRepresentation(vtkVisStrategyRepresentation* vsr);

 // Update all the layers and call \c PrepareForRendering(....) on each. 
 virtual void PrepareForRendering(vtkRenderView* view);

private:

Internal* Implementation; };

class vtkTrackRepresentation : public vtkRenderedRepresentation { public:

 virtual void PrepareForRendering(vtkRenderView* view);

}

class vtkVideoRepresentation : public vtkRenderedRepresentation { public:

 void SetScalarProperty(double value, double range[2]);

 // Can have 0 or many scalar representations. 
 void AddScalarRepresentation(ScalarRepMode);

 virtual void PrepareForRendering(vtkRenderView* view);

}

</source>

Geovis

• Dealing with precision issues

When dealing with large numbers (earth radius is 6357000) in rendering causes issues of Jitter. This is a known problem as the GPU uses single precision floats as the data type. There could be couple of ways this issue can be dealt with.

• • On the fly change of origin
  • Piecewise change of origin
  • Using FO (floating approach as described in this paper link??. http://www.floatingorigin.com/pubs/thorneC-FloatingOrigin.pdf
  • Use lat / lon in radians and height in Km instead of meter.
  • RTC / RTE approaches as suggested here: http://blogs.agi.com/insight3d/index.php/2008/09/03/precisions-precisions/#more-80
  • Use double precision all and cast it to float. (It does help a bit but the problem exists).