Jump to content

 
Photo

Simplifying LiDAR derived building footprints

- - - - -

  • Please log in to reply
3 replies to this topic

#1
jerseysbest

jerseysbest

    Key Contributor

  • Validated Member
  • PipPipPip
  • 51 posts
  • Gender:Male
  • Location:Washington, DC
  • Interests:Road and Mtn Biking, vintage Mustangs, skiing, guitar, and maps of course.
  • United States

I have a whole mess of lidar derived building "footprints" that, at a certain resolution, do represent building density of an area fairly well.

But, those LiDAR buildings up close look terrible; very jagged, sometimes incorporating surrounding trees into the footprint, and are not aligned in a consistent matter

For example:
Posted Image

I was thinking of possibly simplifying what I have in a series of conversions to get a more uniform look, but I'm not sure if its possible or if I'm going about it the right way, or whether its a good idea at all.
  • I was thinking of converting all the LiDAR derived building foot prints to points, and extracting the area of each polygon
  • Then build simple square polygons off the points with the provided area.
  • Then, if possible, align the buildings to the road data. I've done similar things with Arc's Representation tool but I'm not sure it I can do it any other way.

What do you think of this? Good idea? Forget it and stick with what I have? Any thoughts appreciated.

#2
tellett

tellett

    Contributor

  • Validated Member
  • PipPip
  • 11 posts
  • Gender:Male
  • Location:Milton Keynes
  • Interests:Mapping (obviously!!)<br />Photography<br />Music<br />Travelling<br />Sport
  • United Kingdom

It depends what you're looking for in the end I guess. If footprint shape accuracy is the most important
thing then I think you have to stick with what you have because its going to be more accurate than
building shapes from points based on the area field. However, if very simplified representation is ok
then go for the method you suggested, whatever shapes you decide on will look better than those
jagged ones. So I think the end product should really decide it. Thats just my opinion anyway, hope
it helps a bit!!!!

#3
jerseysbest

jerseysbest

    Key Contributor

  • Validated Member
  • PipPipPip
  • 51 posts
  • Gender:Male
  • Location:Washington, DC
  • Interests:Road and Mtn Biking, vintage Mustangs, skiing, guitar, and maps of course.
  • United States

Well, our overall goal is to have accurate building footprints as if they were right from an engineer's planimetrics by way of manual digitizing, importing CAD data from any sources we can get (about 25% of what he have is imported from DWG files), or any way that can accurately represent but also be aesthetically pleasing. This may take months even years, but this is part of a other projects.

In the mean time, we also want the buildings to serve as placeholders for other data associated with the property as well as be part of basemap information for a variety of maps, so shape accuracy isn't paramount right now, but in the end it is, if this data is have any other meaningful value. The LiDAR buildings were seen as a starting point, and we hope to comb through it entirely (I think between 150k and 200k buildings) cleaning it up with multiple people having there hands in digitizing.

Haha, tough call to make I guess, its not the end of the world if I stick with what I have, but I hate looking at all those jagged edges when I'm trying to prepare a simple map.

#4
Lui

Lui

    Master Contributor

  • Validated Member
  • PipPipPipPip
  • 154 posts
  • Gender:Male
  • Location:Ljubljana
  • Interests:Natural looking maps, hillshading, mountain cartography, LIDAR, TLS, data fusion, photogrammetry, singing, poetry, cycling
  • Slovenia

This is quite common problem in LIDAR derived data. I guess that this footprints were derived from holes in LIDAR ground class (LIDAR points classified as ground). I'm using the same method to get "real" not roof based building footprints. LIDAR density, time of year when area was scanned, vegetation density, scanning direction and "squaring" algorithm influence on quality of derived shapes. I'm always using my own LIDAR dataset (we have our own helicopters and LIDAR) and I've find out that a good result can be achived with LIDAR density more than 10pt/m2, scanning in spring when decidious trees have no leaves, scanning with 50% overlap or more, using waveform LIDAR, adeqate ground classification,... But even this kind of LIDAR data doesn't guarantee perfect results. There is always some manual work to do. An idea to get squared footprints from unadeqate LIDAR dataset is to use building classification on LIDAR data and try to get something from this class of points or using DSM (surface model created from 1st echo), nDSM (normalized: DSM - DEM). But this is mainly manual work. I'm sorry that I can't help you more.

Lui




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users

-->