Assigning class codes for LAS files?

Question

I have LAS files with code 0 - never classified. I have no other classes other than this.

How do I assign class codes for my LAS file?

What is the source of this las data set? Sounds like it may be incomplete. — artwork21, Jun 13 '16 at 13:43
This question is broad, but it could also be marked as a duplicate of two complementary existing questions: http://gis.stackexchange.com/questions/132036/classification-of-above-ground-lidar-points-in-arcmap and http://gis.stackexchange.com/questions/188973/classifying-raw-las-data-as-bare-earth-in-arcgis-for-desktop/189458#189458. — Andre Silva, Jun 15 '16 at 11:52

Jeffrey Evans · Answer 1 · 2016-06-14T17:30:33.830

If the class codes field is not populated there is no way to obtain them. During pre-processing the vendor did not classify the points so, you are functionally dealing with a raw point cloud.

Unfortunately, this means that you need to start researching methods for classifying the point cloud yourself. Since you did not provide any context I cannot aim you towards specific resources for classification algorithms. The first step will invariably to classify ground returns then specialized algorithms can be applied to classify certain feature types (eg., buildings). A good software to start with is LASTools, which provides data management and limited classification tools. There is also a free software FUSION that provides some classification utilities. I would not that the ArcGIS LAS utilities for point cloud classification are lacking and one should really look at other software.

score 2 · Answer 2 · edited Apr 13 '17 at 12:34

2

Further to Jeffrey Evans' answer, you could use PDAL to classify the ground points. It's a little harder to get started (on Windows, anyway) but it's free and makes high-volume batch-processing a cinch.

edited Apr 13 '17 at 12:34

Community

1

answered Jun 14 '16 at 16:21

Rob Skelly

3,101
15
14

1

Thanks @Rob Skelly, I was not aware of this implementation of Zhang's morphological filter. I have found that Zhang's algorithm works well in urban environments but, in complex forest structures, it really falls short. There are many other methods available, each with its own advantage. In application it does matter what your question is. If you are after a hydrologically correct DEM one may select a different algorithm than if you want to model forest biomass, where over-smoothing the point could would introduce bias. – Jeffrey Evans Jun 14 '16 at 17:27
@JeffreyEvans I'm curious: in your opinion, which would you say is the best classification algorithm for forested areas? I suppose the nature of the terrain has some impact on the choice? – Rob Skelly Jun 14 '16 at 22:39

score 0 · Answer 3 · answered Jun 14 '16 at 17:32

0

You could try and pull up a land use land cover database of your area where that code 0 is located in. It should be the similar one like this http://landcover.usgs.gov/landcoverdata.php but see if you have one the most recent current land use land cover of your area.

answered Jun 14 '16 at 17:32

PROBERT

4,131
8
49
89

How does this address classified points in the point cloud? There is a standard LAS classification convention that has nothing to do with landcover, but rather the characteristics of the point in the point cloud. There is also an issue of parallax where assigning raster values, with 2D geometry, to a 3D point cloud would introduce significant error. Besides, the landcover data you refer to is at 30m and represents a large scale classification domain whereas lidar data is occurring at submeter and would not be relevant when downscaled to NLCD. – Jeffrey Evans Jun 14 '16 at 17:40
The post was talking about the source that came from drone. So, I was just suggesting to use the land cover land use as a reference because it sound like he was not sure how to code the 0. So, if he use it for that area it might give him a clue what is it there.... – PROBERT Jun 14 '16 at 17:58

Assigning class codes for LAS files?

3 Answers3