Notebooks to generate DSM using reprojection after gridding DSM, and reprojection during pdal processing using smaller tiles.

[ShashankBice]

Hi @scottyhq and @dshean,
I uploaded two notebooks. In this example, the horizontal projection of EPT file is EPSG:3857, and the vertical heights are with respect to the NAVD88 geoid. We found that reprojecting during pdal processing with the compound CRS involved was taking 5 times the time without any reprojection involved. So the first notebook grids the DSM in EPSG:3857, and then uses gdalwarp after the fact specifying the a_srs correctly. The second notebook does the proper reprojection during the pdal processing, using parallelization and smaller tile sizes. There are some differences in the two outputs and I do note that there are seam artifacts caused with the smaller tile size. What do you think would be the best strategy to process them here?

Thanks to @kvenkman for sharing the recipe 😊

Regardless of at what stage we introduce reprojection, I think we also need to understand the different noise filters which are currently turned on by default, as I suspect they might also cause some errors. Starting a different thread to track that down.