[pdal] capturing local geometry descriptors

[Michael Rosen]

Hi PDALers,

I continue investigation into aerial lidar classification prediction. I'm
interested in calculating metrics over a known local volume for each point.
The specific examples I'm considering:

- Normals. Filters.normal allow us to specify how many nearest neighbors go
into the calculation but without consideration to how far away they are.
Does does it make sense to specify a radius in addition to the number of
neighbors? So, perhaps calculate normals based on all points at most 3
meters away:

{

"type":"filters.normal",

"radius":3

}

- Normal variance. HeightAboveGround variance. Along the same lines, I'd
like to augument each point with the magnitude of local normal-variance and
local HAG-variance. Perhaps

{

"type":"filters:variance",

"radius":3,

"dimensions":"NormalX, NormalY, NormalZ, HeightAboveGround"

}



Does anyone have thoughts on how to proceed?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/pdal/attachments/20171020/68bcfd74/attachment.html>