Hi there, I'm working on a project to create a large amount of building footprints. We're planning on giving the building footprints to OSM in the end anyway, so we briefly considered using the Bing imagery in order to digitalize them manually, but to save labor and set up a process that will also work welwell in other areas it would be much better to minimize manual classification if at all possible. That's why we're trying to do the following: 1. Separate non-terrain objects from terrain 2. Separate buildings from other structures 3. Convert 2D building footprints to vectors anand smooth the boundaries
Does anyone have any experience with this kind of operation? I've tried a few things and have a couple more ideas too: - Produce the footprints using a normalized digital surface model. I produced one by subtracting the digital surface model from the digital terrain model (from the data provider) and then classified for all objects over 2 m high. This kept most of the trees and part of our study area is suburban, meaning that there were lots of trees. Most of them coule be eliminated by converting the remaining pixels to areas and then deleting all areas that didn't contain address points, but the results were still at best disappointing. - Another idea I have is to take the LiDAR raw data and try to extract the trees from it, then producing a DSM from the remaining points and normalizing it like above, this time definitely without trees. Since I've never worked with raw data from LiDAR scans before, I imagine this task as being kind of daunting, but nonethless quite interesting. The omnipresent time constraints mak me want to find a way around it, though I don't see how. Anyway, I was just wondering if anyone has any experience here that could help me. If not I'll probably end up scripting an algorithm to separate the trees by the echo they return, if anyone's interested in having it I'd be happy to share. Thanks! Best, Daniel -- View this message in context: http://osgeo-org.1803224.n2.nabble.com/Extracting-building-footprints-from-DSM-DTM-LiDAR-point-cloud-tp6402521p6402521.html Sent from the Grass - Users mailing list archive at Nabble.com. _______________________________________________ grass-user mailing list grass-user@lists.osgeo.org http://lists.osgeo.org/mailman/listinfo/grass-user
