Delivered Ideas

Discuss topics with other Hexagon Geospatial Product pioneers and experts to get the most out of our products.
Showing results for 
Search instead for 
Do you mean 

point cloud classification just for unclassifed points

Status: Delivered
by on ‎03-14-2018 12:21 AM

Some lidar vendors make by default ground etc. classification. For example Swedish NLS provides ground and water. Finnish NLS provides ground and some vegetation. Rest is unclassified. My assumption is that they classify automatically such things they can trust for sure and leave rest empty.


Erdas Imagine Pointcloud classifier makes always a full classification from scratch. So everything is always reclassified. I am not sure can we produce better result for those that someone else has already made. So are we really improving or decreasing result a bit?


I have a decent workaround by using spatial modeller and select only unclassified points to classification. That works pretty OK but fails as not all points are included in process. So I can do the thing I want but only with subset of points which decreases the output quality. What I would like to do is full classification using all points in dataset but new values written just points that had no classification before.


So simple tick in lidar point cloud classifier - classify only unclassified points. That is my request in compressed form.


As an attcahment simple sample model how to work it out but as told it has its weak points too - mainly for decresing the data used and by that compromizing the output quality a bit

Status: Delivered


Delivered as part of the Spatial Modeler 2020 release.

by Technical Evangelist
‎03-14-2018 12:47 PM - edited ‎03-14-2018 01:15 PM
Status changed to: Under investigation


There's an updated version of Timo's model available here, in Spatial Recipes:

on ‎03-15-2018 02:21 AM

And it makes a brilliant job for my datasets and needs, but I agree if preliminiary classification is somehow different it might end up troubles like what happens in SGM based pointclouds if you feed them to point cloud classfier. If there is no multiple echos point cloud classification starts to fail badly.


As a comment to actual request. Adding this feature should be very straighforward. If wished tick is on full point cloud is classified as it is done now and when setting value to point just simply check what class that point already has. If it is 0 or 1 set new value if it is in original data 2-255 do not set new value.


Sounds like fairly simple piece of code - single IF statement more or less Smiley LOL

by Technical Evangelist
on ‎07-12-2018 11:56 AM
Status changed to: Candidate
by Technical Evangelist
on ‎08-06-2019 10:54 AM
Status changed to: Targeted for an upcoming release


Targeted for ERDAS IMAGINE 2020, Q4 2019



by Technical Evangelist
on ‎10-22-2019 07:56 AM
Status changed to: Delivered


Delivered as part of the Spatial Modeler 2020 release.