blast2dem bug: missing pixels when manually specifying output extents

[gberard]

Greetings,

I'm having an issue where manually specifying the output raster extents in blast2dem (-ll, -nrows, -ncols) occasionally results in missing pixels at the edge of the raster.  I believe this to be a bug since it only happens for certain tiles, and when I omit the extent arguments the raster looks fine.

I'm using merged inputs which are tiled in a way that their edges coincide with the desired output extents.  I'm essentially trying to output seamless DEM tiles without having to re-tile the las to add buffers, and it works great 99% of the time until this issue crops up.

Note the missing pixel (left edge, nodata value of -9999) when clipped to extents:

Here's the command.  I don't see anything odd in the output:

D:\Greg_Working\02_Scratch\missing_pixel>blast2dem -keep_class 2 -nbits 32 -step 3
.0 -kill 1000 -i "I:\test\LAS\000442.las" "I:\test\LAS\000441.las" "I:\test\LAS\000474.las" "I:\test\LAS\000473.las" "I:\test\LAS\000443.las" "I:\test\LAS\000472.las" -merged -o 000473_clipped.tif -ll 715050.0 4662900.0 -nrows 150 -ncols 150 -v
WARNING: your license will expire in 27 days.
terrain finalization (using 2d quadtree) to level 7 with empty 0.
have bounding box. first pass not needed.
second pass: populating finalization counter grid ...
second pass: done. created 16384 leaf grid cells and stored 16384 sprinkle points
spfinalizer: needed 3.375 seconds
created parent grid cells (new total is 21845 cells)
third pass: using finalization grid to spatially finalize 7605367 points...
output precision is float ...
third pass: done. stored at most 9220428 chunked points (plus 16384 sprinkle points)
spfinalizer: needed 10.157 seconds
duplicate_points 166
spdelaunay2d: needed 10.161 seconds
read 7605367 points. produced TIN with 7605201 vertices and 15210354 triangles
there were 150 clipped rasters (and 0 unclipped ones)
tin2dem: needed 10.166 seconds
took 13.861 sec to create '000473_clipped.tif'. 

D:\Greg_Working\02_Scratch\missing_pixel>

 

 

The full, un-clipped raster does not show this issue:

D:\Greg_Working\02_Scratch\missing_pixel>blast2dem -keep_class 2 -nbits 32 -step 3.0 -kill 1000 -i "I:\test\LAS\000442.las" "I:\test\LAS\000441.las" "I:\test\LAS\000474.las" "I:\test\LAS\000473.las" "I:\test\LAS\000443.las" "I:\test\LAS\000472.las" -merged -o 000473_full.tif -v
WARNING: your license will expire in 27 days.
llx/lly unspecified. set to lower left bounding box 714597/4.6629e+006.
terrain finalization (using 2d quadtree) to level 7 with empty 0.
have bounding box. first pass not needed.
second pass: populating finalization counter grid ...
second pass: done. created 16384 leaf grid cells and stored 16384 sprinkle points
spfinalizer: needed 1.807 seconds
created parent grid cells (new total is 21845 cells)
third pass: using finalization grid to spatially finalize 7605367 points...
output precision is float ...
lower left x unspecified. setting it to 714600.
lower left y unspecified. setting it to 4.6629e+006.
ncols was unspecified. we set it to 452.
nrows was unspecified. we set it to 302.
third pass: done. stored at most 9220428 chunked points (plus 16384 sprinkle points)
spfinalizer: needed 10.134 seconds
duplicate_points 166
spdelaunay2d: needed 10.137 seconds
read 7605367 points. produced TIN with 7605201 vertices and 15210354 triangles
tin2dem: needed 10.15 seconds
took 12.18 sec to create '000473_full.tif'.
D:\Greg_Working\02_Scratch\missing_pixel>

 

The two commands should be otherwise identical.

The version I'm using is "161114 (commercial)".

I can try to package up some sample data if this is something that needs to be investigated.

Thank you for any input,

Greg