Lidar point density

[AidaAfrooz]

Hello,

What is an accurate enough point density per meter for Lidar data in a low density urban area?

Thanks

[DanPatterson_Retired]

There is no "number".  What are you doing, and where and how precise do you need your results for a desired level of accuracy? How are your measuring accuracy?

[AidaAfrooz]

Hello Dan,

I want to create DEM also calculate building height and if possible generate building foot print automatically instead of digitizing the footprint.

I used Lidar data with point density of approx. 1.5 points per meter (0.8 point spacing) to calculate building height but when I compared the height of some random buildings with the actual height there was an average height difference of nearly 0.9 meters. The height difference of (let's say) less than 0.3 m should be desired in my project since I need to do some view and shadow analysis and 1m difference is a lot.

Thanks

[DanPatterson_Retired]

If that was all the points you have, then you have reached the limit...if not, you can add points to see if things improve, or find out if you have errant points.  What is your height difference relative to in percentage terms is also a consideration?  How accurate is your building height?  Do you know for a fact that it is correct? How was it measured? relative to ground (and where on the ground) or to the top of the foundation?  Just some food for thought. 

[AidaAfrooz]

Yes good to think about all... The actual heights are pretty accurate. I used Earthmine to get the accurate height; however, have checked some heights randomly in Development Applications and compared them with Earthmine which were very close.

I was thinking perhaps there is any range to define accuracy for Lidar point density for example for low density urban areas with X amount of vegetation area... Anyway thanks for your reply.

[Anonymous User · Accepted solution]

Aida. Here are current results per testing

• Highly accurate building footprint extraction and LOD2 (Accurate Roof) Model creation: 8” point spacing
• Good building footprint extraction and descent LOD2 (Accurate Roof) Model creation: 1.7 ft pt spacing
• OK building footprint extraction and descent LOD2 (Accurate Roof) Model creation 2.2 ft or better pt spacing
• Low end: 3 ft pt spacing or better. (LOD2 buildings will be much less accurate looking)
  • Best to just obtain eaveHeight and totalHeight for procedural building modeling.
  • Not sufficient point spacing for building footprint extraction

For Procedural Buildings:

3 ft point spacing or better will suffice. Higher point spacing ensures that with LiDAR classification that trees are not classified as buildings using LP360 or esri's Roof Point Classification GP tool https://www.arcgis.com/home/item.html?id=fe221371b77940749ff96e90f2de3d10

Still recommend 2.2 ft or better at minimum for descent LOD1 classification but 3 ft will still create models.

Should have a tool for these accurate buildings released in the future.

Can also classify LOD1 and LOD2 Buildings from DEM/DTM and DSM rasters.

Pixel spacing at 2ft or better will yield descent results but no-where near as accurate as LiDAR.

Tree point extraction with width and height attributes should be possible with 3ft unclassified LiDAR or better.

We have a tool that can support classified LiDAR but can adapt to unclassified.

Tool can also DEM/DTM and DSM rasters at 3ft pixel size or better in the future as time permits.

[JosephMcGlinchy]

Hi Aida,

To follow up on what Geoff has provided, we also have a workflow that can help generate the building footprints if that data does not already exist by generating DSM and DTM raster surfaces from the point data. What is the classification level of your point cloud data? Do you know if you have ground and building points classified?

Thanks,

Joe

[AidaAfrooz]

Hi Joseph,

I have the ground and buildings as separate classifications (I'm not quite sure if that's what you are asking).  I have 9 classes in the lasd.

May I know more about the workflow you've mentioned?

Thanks,

Aida

[JosephMcGlinchy]

Hi Aida,

If you are confident in the classification of the points in your data, then the workflow which Geoff mentioned should work for you. Feel free to send me a direct message if you are interested in the workflow when your data does not meet those requirements!

Cheers,

Joe