Extract Terrain Surface from Point Cloud¶
Use this dialog to extract the terrain surface from one or more point clouds.
Where do I find the dialog?¶
In the context menu when a point cloud object is selected, choose Tools - Extract terrain surface from point cloud object...
Using the dialog¶
The implemented method is based on the Cloth Simulation Filter (CSF) algorithm. It classifies points as ground points (BE) and object points (OBJ).
The method simulates a simple physical process. First, the point cloud is turned upside down, and then a stiff "cloth" is laid over the inverted surface. By analyzing the interactions between the points in the cloth and the corresponding points in the point cloud, the final shape of the cloth can be determined and used as a basis for classifying the original points as ground points or object points.
A* = Simulated cloth, B = Raw data upside down, C* = Raw terrain surface
Few parameters need to be set in this algorithm. The most important are the terrain conditions and whether you want post-processing of steep areas. The advanced parameters usually do not need to be changed.
With terrain conditions, you determine how the simulation should behave. This depends on the terrain type, and you can choose between flat, normal, and rough terrain. If your area contains all these types, it is natural to choose rough. If this does not give the desired result, you can consider splitting the point cloud.
For steep slopes (cliffs, escarpments), the CSF algorithm can produce relatively large errors because the simulated "cloth" may be too stiff. This can be solved with post-processing of steep slopes. If there are no steep slopes in the point cloud, you do not need to use this option.
Info
A detailed description of the theory and algorithm can be found in the article: An Easy-to-Use Airborne LiDAR Data Filtering Method Based on Cloth Simulation.
There is also an option for inverse simulation. This can be used if you want to find the top of a forest area to calculate volume. For such an example, you can choose Flat terrain and Post-processing of steep slopes to produce a smooth top surface of the forest, while, for example, roads remain unaffected.
Advanced parameters¶
The resolution refers to the grid size of the "cloth" laid over the terrain. The larger the resolution you specify, the coarser the result. How are the points for the cloth selected? Within each grid cell, the highest point is chosen.
The algorithm produces a grid model (mesh) for the "cloth"—the simulated terrain surface—which all points are checked against. The threshold distance controls the final classification of points as ground or object points, based on the distance between the points and the simulated terrain.
As mentioned above, the Cloth Simulation Filter (CSF) algorithm will classify the points as ground points and object points.
Example
Unfiltered point cloud
Classified as ground points
Point adjustment¶
There is also an option to adjust points, meaning to adjust the ground and object points to the "cloth". With this option, you get a smoother result.
The adjustment can be done in two ways, as illustrated in the figure below:
- Move existing points
- Insert simulated points
A* = Simulated cloth, B = Original points, C = Move existing points, D* = Insert simulated points
Ground points with adjusted object points
With the Insert simulated points option, you get a result with a resolution equal to the grid density of the "cloth". With Move existing points, interpolation is performed in the grid, and you retain the number of points in the result.
Result options¶
There is also an option to fill holes, meaning areas without points (ground or object points), as shown in the figure above.
You can get the result separated into two point clouds: one for ground points and one for object points.
Limitations¶
The CSF algorithm also has limitations. Some object points (OBJ) may be incorrectly classified as ground points (BE) when working with very large, low buildings. In addition, the CSF algorithm cannot distinguish objects that are connected to the ground (for example, bridges).