Cloud Compare is an open source software for point cloud processing and editing.
There’s an interesting plugin to filter point clouds (from photogrammetry and laser scanners) and isolate the points of the ground from those that are not.
It’s called “CSF Filter” and you find it among the plugins.
Its principle is very interesting!
Take a sheet (cloth means fabric) and place it on top of a 3D point cloud, with trees, houses and cars in it.
The surface of the sheet represents a DSM (Digital Surface Model).
Now turn the 3D model upside down and let the sheet drop again on the inverted cloud.
The surface of the sheet is now a DTM (Digital Terrain Model).
The difference between the two surfaces that have been found, allows you to classify the point cloud by extracting just the ground elements.
Actually other parameters also come into play and it’s not so simple to go deeper in te, but I think the idea behind this algorithm is really interesting.
CSF Filter plugin is based on some settings.
The default ones may not give good results.
If you open the advanced parameters window and gradually change (lowering them) “Cloth Resolution” and “Classification Threshold” (at he very beginning leave the “Interations” parameter unchanged at the default value), you may be able to distinguish the soil from vegetation and anthropic elements.
The plugin works best if the vertical elements to be “removed” from the point cloud are defined in their elevated parts.
A lot is made by the nature and quality of the point cloud itself.
Of course it doesn’t have to be an impenetrable continuum.
Removing the vegetation in order to reach the ground in a point cloud of a dense forest, after processing nadiral pictures from above with photogrammetric techniques, is destined to fail miserably.
Other situations could be tricky too.
But sometime the CSF plugin works well.
Bare in mind that an automatic process like this one should always be kept under control
Do you have experience with automatic procedures for the classification of point clouds, by photogrammetry or by laser scanner?
In order to share the tools that i use let me say that to classify a point cloud i religiously use LiDAR360 algorithms.
Feel free to fill the comments below with your “story”.
I will really appreciate and it will really be useful for others.