Step size for large barns in rural terrain

23 views
Skip to first unread message

Andy Ford

unread,
Jan 20, 2017, 9:11:32 AM1/20/17
to LAStools - efficient tools for LiDAR processing
I'm new to LAStools and new here.

I have a large data set covering a rural area with gently undulating terrain, but also interesting local geomorphological details I wish to preserve. There are villages and towns, which I've so far efficiently removed by classifying with lasground_new and default values. It mostly works great, except for large farm buildings, i.e. barns. There is also the odd airfield or two with large hangars which refuse to be removed! So now I'm in the realm of experimenting with step sizes. The barns and hangars range in width from 30-60m and length 100-200m. If they were square I can see how simply increasing the step size would do the job, but what about rectangular buildings, especially very elongate ones? Do I aim to bridge the shortest dimension or the longest? I also want to ensure that by increasing the step size so much I don't sacrifice too much small terrain detail...

Any advice is appreciated.... Thanks in advance!

Terje Mathisen

unread,
Jan 20, 2017, 8:34:56 PM1/20/17
to last...@googlegroups.com
You can try to run everything twice:

First you process with -town and lasclassify to detect all major
buildings, then you re-run lasground_new with '-wilderness -ignore_class
6' in order to ignore all points previously classified as buildings.

Terje

--
- <Terje.M...@tmsw.no>
"almost all programming can be viewed as an exercise in caching"

carlisle haworth

unread,
Jan 20, 2017, 8:35:18 PM1/20/17
to last...@googlegroups.com
Hi Andy,
I ran into a similar situation when working with a bunch of huge railroad depots. There were massive warehouses and also small sheds, not to mention all the rail cars themselves and the occasional vegetations. One solution that worked pretty well for me was to actually perform two separate ground classifications on the same dataset. My datasets were completely unclassified, so in the first round I would run a ground classification using more of a rural stepsize (looking for smaller buildings and features) so you end up with all the points now being either ground or non-ground with none of them being unclassified. This tended to grab the sheds, cars, vegetation, etc. but would not do a great job of identifying the large warehouses. Next, I would take everything that had been classed as ground in the first step, and change it back to unclassified. Then I would run another ground classification on it using more of an urban step-size, and this would find the big buildings no problem. So it's an unusual approach, but worked great and resulted in the least amount of manual touch up.

Carlisle

Reply all
Reply to author
Forward
0 new messages