[pdal] DEM Grids - p2g bounding box

[Matthew Hanson]

Pete - this looks like exactly what I need!   At least on the p2g side.
Now, what's the best way to incorporate this into the PDAL plugin?

Andrew, the problem with setting the bounds after the fact is that the
calculated grid could be shifted by fractional pixels from the desired
one.  So you really need to be able to specify the starting grid so that
your end result has properly interpolated at the pixel centers you'll want
to use.

Matthew Hanson
Applied GeoSolutions
(603) 659-3363 x91
http://appliedgeosolutions.com
mhanson at appliedgeosolutions.com


On Tue, Mar 24, 2015 at 11:48 AM, <pete.gadomski at gmail.com> wrote:

> Matthew,
>
> We added the ability to specify a user-defined grid in Points2Grid a few
> months ago, thanks to work by Tristan Dyer.
> https://github.com/CRREL/points2grid/commit/b2ada6abfc85560ec10fb4f30333a1bf95be9025.
> Is that what you had in mind?
>
> -Pete
>
>
>
> On Tue, Mar 24, 2015 at 10:41 AM, Matthew Hanson <
> mhanson at appliedgeosolutions.com> wrote:
>
>> We've run into another little snag trying to create DEMs.    I made a
>> commit last month to fix the half pixel offset in p2g output:
>>
>> https://github.com/PDAL/PDAL/commit/ea67d14f43b09e591ef94c7f03ef75d64dc786ed
>>
>> Which is correct, the output is not correctly aligned to how it's doing
>> the gridding, however the gridding itself isn't really ideal.   There are 2
>> problems:
>>
>> 1) p2g uses the upper left point (min x, max y) as the *center* of the
>> top left pixel.   So this means that there are going to always be fewer
>> points to choose from when on a corner of edge of the resulting raster.
>> How many fewer points depends on the search radius.    While this depends
>> on the output of the raster (e.g., min, max, idw) this might not be
>> noticeable at all, but there are certainly some edge effects from this.
>> Granted if p2g used half a pixel in from the minx/maxy point it could still
>> suffer from this problem if given a search radius > 0.5 pixel, but I think
>> we can agree that the using the top left available point as the center of
>> the pixel is not the best way.
>>
>> 2) Related to the above problem is the more general one that the raster
>> bounding box is completely reliant on the input point cloud.    When using
>> point clouds that have been tiled, or otherwise have specifically chosen
>> bounds (but perhaps points do not go all the way to the edge) the output
>> raster may not have the same bounding box.   However, the even greater
>> problem is that if you create a DTM from ground points and a DSM from
>> non-ground...well, that top left point is going to be different between
>> those and therefore you get two DEM rasters that are potentially on
>> completely different grids.
>>
>> What is really needed, and would solve both issues in a general way, is
>> to be able to specify the grid that should be used (really this just means
>> being able to give it an exact top left coordinate).    This would allow
>> alignment of DTM and DSM without requiring postwarping (really not
>> desirable).
>>
>> It looks like this would mostly be a change within p2g to take in a
>> starting x,y coordinate, and in PDAL the addition of another argument to
>> pass to OutCoreInterp.
>>
>> Does this seem like desirable behavior?
>>
>>
>>
>> Matthew Hanson
>> Applied GeoSolutions
>> (603) 659-3363 x91
>> http://appliedgeosolutions.com
>> mhanson at appliedgeosolutions.com
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/pdal/attachments/20150324/b550f8b5/attachment.html>