Thanks for the great ideas. Those sound promising. I'll give them a whirl and report the results!<br><br>Mark<br><br><div class="gmail_quote">On Mon, Jun 2, 2008 at 4:56 AM, Hamish <<a href="mailto:hamish_b@yahoo.com">hamish_b@yahoo.com</a>> wrote:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">Mark wrote:<br>
> I suspect the intensity data may be better for house extraction,<br>
> because the roofs are mostly all uniform in intensity clusters<br>
> values, where with RGB the roof shingles are far to varied for it<br>
> to classify them cleanly without other classes of non-roofs appearing<br>
> all over the map in scattered clusters.<br>
<br>
Hi,<br>
<br>
an idea about extracting roofs from LIDAR data- create a slope map from r.in.xyz DEM and run i.cluster etc. on that to extract areas of constant slope. Then threshold that to remove small area noise* (say under 25 sq-m) and large areas (roads, etc).<br>
<br>
[*] in vectors this would be 'v.clean tool=rmarea', not quite sure about cluster data, maybe r.grow/r.buffer the null cells then back the other direction and use that as a mask??<br>
<br>
Not sure how to separate "big box" Walmarts from parking lots, maybe if you set high slopes to NULL you can turn flat areas into islands, then test each island (r.to.vect areas?) for height departure from the surrounding DEM?? perhaps something with r.neighbors?<br>
<font color="#888888"><br>
<br>
Hamish<br>
<br>
<br>
<br>
<br>
<br>
</font></blockquote></div><br>