RE: [LAStools] Classification
Chris Ryan
Hi Lloyd,
I have done a bit of work in this field (also for hydrologic modelling purposes) and have had some success. I have found the biggest problem to be the step size parameter in lasground. In my projects, the building size often varies widely even in the same tile and thus no step size works perfectly. If the step size is too small, large building can be missed (or have holes in the middle) and if it’s too large, you can get misclassification of bridges, road abutments etc as buildings. Like you, I do try to filter out pulses with multiple returns and I also filter out points based on height (which I stored in a user data field using lasheight) as I find a lot of misclassified building points have a relatively low height value compared to genuine building.
As far as impervious areas go, I used GDAL to create a pseudo-NVDI image using the red band of the air photo and the intensity grid from the LiDAR. I was able to classify roads reasonably successfully based the NVDI < 0.5. You can also come up with a couple of vegetation categories based on their ‘greenness’ using the NVDI. Clear water can be classified pretty well based on very low intensity values and tree canopy can be very well classified based on multiple returns.
Anyway, I hope this helps. I haven’t been able to achieve the accuracy I want from this method yet but I’m getting closer. Any advice from the community would be appreciated.
Martin, for Christmas, I would like a version of lasground that doesn’t require a manual stepsize J
Regards,
Chris
From: last...@googlegroups.com [mailto:last...@googlegroups.com] On Behalf Of lloyd.fis...@gmail.com
Sent: Thursday, 25 October 2012 10:52 PM
To: last...@googlegroups.com
Subject: [LAStools] Classification
Hi
I am new to both this group and processing LAS data. Essentially I want to Identify roof's and impervious area's. I was wondering whether anyone can assist with what parameters/ method I should use. I am hoping to use the data to generate a more accurate stormwater model for a small catchment in South Africa as part of my PhD looking at various stormwater management options in South Africa. The catchment is relatively high density residential with a number of blocks of flats, and then tightly packed single story houses on the slopes of Table Mountain. I have followed the following procedure:
- LAS Ground
- LAS Height
- LAS Classify (and tried varying the parameters - but not exactly sure of how best to))
but whatever I do I dont seem to get a good classification, sometimes in one area but not in another.
I have tried ignoring ground points and only working with heights. I recently tried creating a grid based on number of returns( and this was by far the best at Identifying high vegetation) So at the moment the best method I can see is to run LAS height, then remove all points with Higher number of returns and then treat all remaining points as the buildings. This will be relatively useful, but I doubt its the best method? It also wont help with the Impervious ground areas, but I haven't yet tried working on those.
If anyone has any advice it would be greatly appreciated.
Regards
lloyd
--
Download LAStools at
http://lastools.org/
Visit the LAStools group at
http://groups.google.com/group/lastools/
Be social with LAStools at
http://www.facebook.com/LAStools
http://www.twitter.com/LAStools
Jessica Block
Don Marsh
Chris Ryan
Hi Don,
You can find a paper on our website http://csse.com.au
Just go to the Publications widget on the right and click on Using LiDAR Survey for Land Use Classification (2013)
Cheers,
Chris
From: last...@googlegroups.com [mailto:last...@googlegroups.com] On Behalf Of Don Marsh
Sent: Wednesday, 18 June 2014 12:02 PM
To: last...@googlegroups.com
Subject: Re: [LAStools] Classification
hello Chris, Yes. I am interested also in the GDAL NDVI work you've done. That's great - fusing imagery with LiDAR. Can you share more details?
On Monday, October 29, 2012 2:44:44 AM UTC-4, cr...@csse.com.au wrote:
Hi Lloyd,
I have done a bit of work in this field (also for hydrologic modelling purposes) and have had some success. I have found the biggest problem to be the step size parameter in lasground. In my projects, the building size often varies widely even in the same tile and thus no step size works perfectly. If the step size is too small, large building can be missed (or have holes in the middle) and if it’s too large, you can get misclassification of bridges, road abutments etc as buildings. Like you, I do try to filter out pulses with multiple returns and I also filter out points based on height (which I stored in a user data field using lasheight) as I find a lot of misclassified building points have a relatively low height value compared to genuine building.
As far as impervious areas go, I used GDAL to create a pseudo-NVDI image using the red band of the air photo and the intensity grid from the LiDAR. I was able to classify roads reasonably successfully based the NVDI < 0.5. You can also come up with a couple of vegetation categories based on their ‘greenness’ using the NVDI. Clear water can be classified pretty well based on very low intensity values and tree canopy can be very well classified based on multiple returns.
Anyway, I hope this helps. I haven’t been able to achieve the accuracy I want from this method yet but I’m getting closer. Any advice from the community would be appreciated.
Martin, for Christmas, I would like a version of lasground that doesn’t require a manual stepsize J
Regards,
Chris
From: last...@googlegroups.com [mailto:last...@googlegroups.com] On Behalf Of lloyd.fis...@gmail.com
Be social with LAStools at