A Portal to High-Resolution Topography Data and Tools

  

Data    Tools    Site
Follow OpenTopography on Twitter

Find Us on Facebook

Tools for Lidar Point Cloud Filtering / Classification    

Posted on Tue, March 08, 2011 by Chris Crosby in point cloud • Software

A relatively common question from OpenTopography users is how they can filter or classify a lidar dataset that was delivered by the data provider without ground (bare earth) returns differentiated from vegetation returns. The B4 lidar dataset, which covers the southern San Andreas and San Jacinto faults is a good example of a dataset where the lack of classification can be problematic for users, especially those working at higher elevations where vegetation can be dense.  In other cases, the classification was done hastily or not as well as a user would like, and thus re-classification is needed.  A good example here are the 2010 post-Haiti earthquake data which were processed rapidly in order to quickly distribute them to users.

Most lidar vendors use expensive black box commercial software to classify data - TerraScan is the industry standard - not readily available to your average Earth scientist.  This of course raises the question: For users who don’t have access to a commercial product like TerraScan, what other tools are available for performing lidar point cloud classification?

This week, someone on the libLAS developers list asked essentially that very question.  The question triggered a series of posts by other list members on software currently available to classify lidar data and develop bare earth terrain models.  Some of the suggestions were tools that I was already aware of (including some already registered to our new OpenTopography Tool Registry), while one or two were new to me.  Below is a compilation of suggestions provided by the libLAS-devel listserv members.

Tools for Lidar Classification / Filtering:

  • MCC-LIDAR - An open source command-line tool for processing discrete-return LIDAR data in forested environments. It classifies data points as ground or non-ground using the Multiscale Curvature Classification algorithm developed by Evans and Hudak, 2007

  • GRASS GIS - Open source GIS software.  Includes a suite of tools related to lidar data processing as discussed in this GRASS lidar wiki entry.  Currently doesn’t offer direct support for point cloud data in LAS format.

  • BCAL Lidar Tools - Open source tools developed by the Idaho State University Boise Center Aerospace Lab in IDL as a plugin for the ENVI software package.  Includes a Height Filtering tool optimized for open rangeland (sagebrush) vegetation developed by Streutker and Glenn, 2006.

  • SAGA GIS - Open source GIS package.  “SAGA includes several tools to manipulate the point cloud, e.g. an attribute calculator, reclassifier, subset extractor and many other methods like gridding and interpolation. There is also (a grid based) bare earth filter, adapted from Vosselman (2001).”
What other options are out there?  The libLAS list discussion emphasized open source tools, but I know there are other commercial options (e.g., Merrick’s MARS), that also offer point cloud classification capability.  If you are aware of tools not listed above, please leave a comment.  I have varied experience with these tools, so your thoughts and recommendations are also welcome in the comments. TerraScan is presumably the industry standard because when applied by a skilled operator the classification results are generally good.  Another issue inherent in all lidar data processing tasks is software that has the scalability to handle the massive data volumes typically encountered.  I assume that TerraScan does relatively well on this front, but it is an issue to also keep in mind when looking at the alternatives listed above.

User controlled point cloud classification (or re-classification) is a feature that we’d like to offer through OpenTopography, and one or more of these open source tools may be a viable option for integration into OT in the future.


Bookmark and Share

I recommend the use of FUSION - it’s free and very useful:
http://forsys.cfr.washington.edu/JFSP06/lidar_&_ifsar_tools.htm

By: Greg on 03/14/2011

Thanks Greg - I’d forgotten that Fusion does ground modeling.  We’ve used it for point cloud viewing in some training classes (e.g., http://www.opentopography.org/index.php/resources/short_courses/09SCEC_course/).

Instructions on how to use FUSION for point cloud filtering are buried in the “Creating a Fusion Project Part 2” tutorial, located here: http://www.fs.fed.us/eng/rsac/fusion/

By: Chris Crosby on 03/14/2011

Another tool set that has been around a while is the Airborne LIDAR Data Processing and Analysis Tools
http://lidar.ihrc.fiu.edu/lidartool.html

Provides a number of different filtering algorithms including a version of the Adaptive TIN method described by Axelsson (2000).  In the paper below, the author states that “a version” of this method is implemented in TerraScan.  Don’t know what differences exist- especially since the time of writing, now 10 years ago.

Axelsson, P., 2000. DEM generation from laser scanner data using adaptive tin models. International Archives of Photogrammetry and Remote Sensing, XXXIII, Part B3, 85-92

As for commercial applications, Tiffs advertises robust filtering and processing- well suited for forested settings.  I believe there is a one-week, fully-functioning demo license available for evaluation.

Tiffs
http://globalidar.com/default.aspx

By: Russ on 03/18/2011

Chris,

Thanks for the shout-out - much appreciated!

Yes, MARS does support both automated and manual LiDAR data classification tasks. Users can utilize any or all of the 25 filter algorithms to create a custom macro script for specific classification tasks. The current filter algorithms generally operate on the attributes of a LiDAR dataset - returns, intensity, scan angle, timestamp, etc.

Internally, we use this method to automate the classification process before our technical staff performs manual updates. Typically, we can process tens of thousands of tiles using this automated approach, which utilizes multi-core processing to speed up the work.

Going forward, we plan on further developing our feature extraction capabilities and refining our bare-earth extraction process.

In terms of pricing, MARS is a good value for its price ($4,995 for Version 7.0), especially when compared to Terrascan (needs Microstation to run!) and other commercial offerings.

I’ll be glad to provide further insight in to LiDAR classification to those individuals who have interest in doing this type of work.

By: Bill Emison on 03/23/2011

VRMesh Survey provides an intelligent solution for automatic point cloud classification and bare-earth extraction. It automatically and accurately classifies LIDAR point clouds into ground, vegetation, building, and others. More than 90% identification jobs will be done in a one-click process. Manual adjustments can be easily performed in a cross section view. It also provides comprehensive functionalities to generate accurate triangle meshes.

By: Chao on 03/29/2011

Thanks everyone for your comments - lots of great suggestions that I plan to investigate further.  This post and the comments are becoming a great resource. 

Just a reminder for those of you involved with developing lidar-oriented software, you are welcome to register your tool(s) with the OpenTopography tool registry: http://opentopo.sdsc.edu/gridsphere/gridsphere?cid=contributeframeportlet&gs_action=listTools

Contribution (http://opentopo.sdsc.edu/gridsphere/gridsphere?cid=contributeframeportlet&gs_action=addTool) requires an OT login and it should only take a couple minutes to to fill out the required info.

By: Chris Crosby on 03/29/2011

Check out the tools at http://www.cs.unc.edu/~isenburg/lastools/
Just updated Dec 13 2011, includes classification tools.

By: Michael P Sartori on 12/15/2011



Commenting is not available in this weblog entry.