[GRASS-user] Problems with lidar tools

GRASS GRASS user
1

Hello,

I'm having major problems with the Lidar tools. I've got a dataset with
5.5-7.0 points per m² and have run v.edgedetection on the original point
cloud containing all points with a the spline step parameters set to 0.6 and
the other parameters set as outlined in "Inverse calibration of lidar
filterimg parameters" by Brovelli & Lucca. Nonetheless, the edges detected
don't seem to have a whole lot to do with actual edges in the scene,
buildings look like they're completly ignored. Another fact is that I can't
see the data in nviz, it claims that there are no points in the region, even
though I've set the region settings to match the LiDAR data set.

More confusing, I can tell GRASS to show the points with random colors
according to their classes, but I can't see the classes themselves in the
attribute table, so I can't extract them thematically or control how they're
shown. If I build topology again for that classes, I can open the attribute
table, but it only has two fields: ID and Interp. Can't figure out what they
mean, and v.lidar.growing doesn't change the map at all. What's going on??

Thanks a bunch,
Daniel

--
View this message in context: http://osgeo-org.1803224.n2.nabble.com/Problems-with-lidar-tools-tp6418988p6418988.html
Sent from the Grass - Users mailing list archive at Nabble.com.

2

On Mon, May 30, 2011 at 2:57 PM, LeeDaniel <Lee.Daniel.1986@gmail.com> wrote:

Hello,

I'm having major problems with the Lidar tools. I've got a dataset with
5.5-7.0 points per m² and have run v.edgedetection on the original point
cloud containing all points with a the spline step parameters set to 0.6 and
the other parameters set as outlined in "Inverse calibration of lidar
filterimg parameters" by Brovelli & Lucca. Nonetheless, the edges detected
don't seem to have a whole lot to do with actual edges in the scene,
buildings look like they're completly ignored. Another fact is that I can't
see the data in nviz, it claims that there are no points in the region, even
though I've set the region settings to match the LiDAR data set.

If I understand the Brovelli & Lucca procedure correctly, there are no
hard rules to determine the settings for v.lidar.egedetection, instead
it is trial and error (iterative calibration against a reference). You
may have to experiment with different settings until the edges of
buildings are properly detected.

About nviz, I guess nviz needs vectors with topology for display, and
the output of v.lidar.edgedetection does not have topology, you may
need to run v.build manually.

More confusing, I can tell GRASS to show the points with random colors
according to their classes, but I can't see the classes themselves in the
attribute table, so I can't extract them thematically or control how they're
shown. If I build topology again for that classes, I can open the attribute
table, but it only has two fields: ID and Interp. Can't figure out what they
mean, and v.lidar.growing doesn't change the map at all. What's going on??

In the output of v.lidar.edgedetection, the classes TERRAIN, EDGE,
UNKNOWN are not in an attribute table, they are only coded by cats.
You can add a table to layer 1 with columns cat and label and populate
the table with label = TERRAIN for cat = 1, label = EDGE for cat = 2
and label = UNKNOWN for cat = 3.
The table with the fields ID and Interp is attached to layer 3.

v.lidar.growing adds categories to layer 2, but no attribute table.
Again, you would need to add a table manually. And there is probably a
bug in v.lidar.growing, needs (time for) testing.

HTH,

Markus M