Closed rosepearson closed 1 month ago
An example of the points to interpolate:
Notes on interpolation approach I will try cubic / RBF, but I also want to explore a better patched approach using a KD tree to define each patch based on the points in + 1/2 radius away from the block with a warning noting that this approach will only give smooth results for relatively densely space points.
Would be built on:
Another approach is take (say) 15 nearest to each point and interpolate using a spline there..
Allow the set foreshore to zero to be turned off for when there is a 'good' offshore dataset.
Have done a fair bit of experimentation. Main notes are:
Images (LiDAR grey vs offshore points white)
Have created a test case. unresolved questions exist around what to do when edge points are being used (e.g. LiDAR data along the coast is used) and this data is much closer than the sparse offshore data. Currently only taken the mean of these values to limit artefact from different density of data being considered together.
Test with linear interpolation and the edge points being used.
Test with rbf interpolation and edge points being used
Test with RBF interpolation and no edge points being used - thin plate spline (try other kernals)
Test with RBF interpolation and no edge points being used - linear kernals - not finite support so interpolation of some points
Re-expose support for offshore interpolation of sounding points
Currently experimenting with different interpolation approaches (linear cubic and Rbf). Will need to expose additional parameters as can't infer appropriate search radius for radius based KDTree search for near points from the input data.