TGIS_InterpolationSplines class

DK11 for Delphi | GisInterpolation.TGIS_InterpolationSplines | Classes | Constructors | Methods | Properties

Implementation of the Completely Regularized Splines (CRS) interpolation method.

Available also on: .NET | Java | ActiveX | Python.

// Delphi
type
  TGIS_InterpolationSplines = class( TGIS_VectorToGridAbstract )
  end;

// C++ Builder
class PASCALIMPLEMENTATION TGIS_InterpolationSplines : public TGIS_VectorToGridAbstract
{
};

→ TGIS_BaseObjectDisposable → TGIS_VectorToGridAbstract

The principal of the Completely Regularized Splines interpolation method is to minimize the overall surface curvature. It produces the smoothest results among the vector-to-grid interpolation techniques.

There are two different scenarios for generating a splines interpolated grid:

1. based on a Z, M, or Z+M coordinate value:
   • set the Coordinate property to the desired option
   • pass an empty string as the field name to the Generate method
2. based on an attribute field value:
   • pass the field name to the Generate method
   • with the above, the Coordinate property is neglected

The Tension parameter is a smoothening factor - the smaller the tension the less rapid changes in the interpolated values. For best results it must be experimentally adjusted for each input dataset. The most common values of Tension (that are good to start with) are in the range of 0,00001 to 0,1.

Use a windowed version of the algorithm (Windowed=True) for large datasets (thousands of points and more). For up to a few hundred points, the non-windowed (full sample) algorithm is efficient enough and gives best possible results. The Radius property is the size of the window in map units. The MinPoints/MaxPoints properties define the minimum/maximum number of sample points necessary/taken to interpolate a grid value.

• smoothest results of all the interpolation methods (if this is the goal)
• interpolated values exceed the range of minimum to maximum values of the input data

• for many applications (like terrain models), the results are too smooth and seem unnatural
• works badly on input data with extreme value changes at short distances (because it amplifies the differences)

var
  src : TGIS_LayerVector ;
  dst : TGIS_LayerPixel ;
  vtg : TGIS_InterpolationSplines ;  
  ext : TGIS_Extent ; 
begin
  // get the source layer (GIS is TGIS_ViewerWnd)
  src := TGIS_LayerVector( GIS.Get( 'points' ) ) ;
  ext := src.Extent ;
 
  // create the destination layer (grid, 200x200 pixels)
  dst := TGIS_LayerPixel.Create ;
  dst.Build( True, lsrc.CS, ext, 200, 200 ) ;
  dst.Name := 'Heatmap' ;
  dst.Params.Pixel.GridShadow := False ;
 
  // create an instance of the splines interpolation engine
  vtg := TGIS_InterpolationSplines .Create ;
  try
    // attach event to track the progress
    vtg.BusyEvent := doBusyEvent ;
 
    // set the tension parameter
    vtg.Tension := 0.0001 ;
 
    // initiate the grid generation process for the
    // POPULATION attribute field
    vtg.Import( lsrc, ext, 'POPULATION', ldst, ext ) ;
  finally
    vtg.Free ;
  end ;  
 
  // add the grid to the viewer
  GIS.Add( dst ) ;
 
  // update the viewer
  GIS.FullExtent ;
end ;

• Interpolation sample