Point Cloud Classification Settings

Hi All. I am new to the group and started using GM a few months ago for the sole purpose of cleaning and classifying point clouds produced from Pix4D. While cutting sections and classifying manual has been very successful it is time consuming and I am sure there is a better way. 

I am trying to quickly get rid of all the noise above and below the road surface shown in the attached sample section. I have tried Auto-Ground and Auto-Noise with various settings and keep coming up with bogus classifications. Could anyone help me to better understand a work flow to clean something like this up?

Also, I have looked all over the place trying to find a good reference for the GM settings and really haven't found much that helped. If I missed something, please let me know. I am not above doing my own research to solve my problems. I just hadn't found something that helped yet.

Thank you all very much in advance!


  • NatsNats Posts: 16
    Dear Nickm1622.

    Gm is cannot handle the Point cloud properly, as per auto classification it will not perform as what your are looking for.
    I would suggest you to use Terra scan which helps in classifying the Noise (through the macros) i,e Above ground/and below ground.


  • I have had some luck processing UAV imagery generated point clouds with the LiDAR module, but am still working to to 'good' results.  Is it the consensus of this forum that the GM LiDAR module is not appropriate for Drone generated point clouds?  That is a shame if it is true.  If not, is there someone with some good news that can chime in with suggestions?  
  • KatrinaKatrina Posts: 22
    We have been continuing to make improvements to the LiDAR module automatic classification tools that improve how they handle UAV collected LiDAR or photogrammetric point clouds. The most recent improvements have not been through the full QA process yet, but are available in the daily build of the application: http://data.bluemarblegeo.com/downloads/global-mapper/dailybuilds/

    We have many users successfully working with UAV collected lidar data or photo-derived point clouds. Here is one case study: http://www.bluemarblegeo.com/docs/case-studies/dronemapper-case-study.pdf ;

    Here is the help file information about the tools: http://www.bluemarblegeo.com/knowledgebase/global-mapper-18-2/#Lidar_Module/Automated_Lidar_Analysis_Tools.htm ;
    We also just completed a Lidar Webinar series that talks more about these settings: 
    Upon request to geohelp@bluemarblegeo.com we can also provide additional information about the classification algorithms used, or suggest settings to refine the automatic classification given a data sample. 

    The recommended workflow for a completely unclassified point cloud would be to first perform noise detection using the standard deviation setting. Then perform the ground classification. The settings in the 'removal of likely non-ground' section are a pre-filter, that will remove points from consideration based on changes in height. The latest changes also include a setting here for the 'Maximum Building Width', which was a change made to help with cases like large commercial buildings, where part of the flat roof may become classified as ground. 
    If you are having trouble with the values, we would recommend running the algorithm on a small section of the data using the Specify Bounds option. Tweak the setting on that sample area to get a good classification result, and then apply the same values to the whole point cloud. 

  • Thanks for response, Katrina.  I have seen the excellent webinar.  The issue seems to be with the higher point densities of the photo generated point cloud.  I will continue to work with smaller sections.  I don't know if there are any recommended settings for such a file as a starting point.  Using '1 point' spacing for the gridding seems to tight, and 1 or 2 meters may be to large.  There are so many settings, it is a little daunting. I know there is personal training available; however, is there a chance to just talk with someone on the phone for a short conversation?  BTW, the link for the case study above did not work.  
  • MykleMykle Global Mapper User Posts: 397Trusted User
    When you open the case study and receive the file-not-found screen, edit the address to remove the characters after ".pdf" and the file will then open fine.
Sign In or Register to comment.