07. Terrain Analysis - VUKOZ-OEL/3d-forest-classic GitHub Wiki

There are two automatic methods for terrain processing. None of these methods are perfect, so it is advised to use both of them to combine the advantages of their outputs. There is also a possibility of manual adjustment of the results.

Terrain from Octree

Go to Terrain → Terrain from octree. In the dialog window select the base cloud that should be processed. Enter the name of the new terrain cloud, the name of the cloud where the rest of the data (usually vegetation) will be stored, and set the resolution (length of cube edge in cm). The input cloud is divided into cubes. Cubes containing points and having the lowest z-value are considered to be the "ground cubes". The terrain is then defined by the points in these ground cubes. This means that all the points in ground cubes are considered to be terrain and contrary to voxelization there is no reduction of points and no generalization. This method is more laborious for manual post-processing. However, the results are more detailed and built by original points. HINT: If there are points that belong to the terrain within the vegetation cloud, i.e. points connecting trees after the segmentation, it's recommended to use a higher resolution.

Terrain from Voxels

Go to Terrain → Terrain from voxels. In the dialog window select the base cloud to extract the terrain. Enter the name of the new terrain cloud, the name of the cloud containing the rest of the data, and set the resolution of voxelization (length of voxel edge in cm). Appropriate resolution to gain relevant results is below 50 cm. The input point cloud is reduced into voxels and centroids of voxels (i.e. points with averaged coordinates from all original points of individual voxels). Voxels with the lowest z-values define the terrain. Increasing voxel size also rise the overestimation of terrain height (z-value of centroids is affected by more points which lie above the terrain). However, as the resulting cloud contains fewer points than the cloud produced by Terrain from the octree also manual post-processing is usually faster.

Statistical Outlier Removal Filter

Both terrain cloud from voxels or form octree may be filtered using statistical outlier removal. Go to Terrain → Statistical outlier removal. In the dialogue window select terrain cloud set. Name the cloud with removed points, and set the number of neighbors for computing mean distance and standard deviation. Points with a mean distance longer than standard deviation are removed. For more details see the PCL documentation.

Radius Outlier Removal Filter

Applying radius outlier removal filter to terrain cloud is possible via Terrain → Radius outlier removal. In the dialogue window select terrain cloud set. Name the cloud with removed points, set the radius, and the number of neighbors. Points with fewer neighbors in a given radius are removed. For more details see the PCL documentation. Tips: If too many or too few points were removed by filters use the cloud merge feature and repeat the procedure. For the best result combination of both filters is recommended. Note that different voxel sizes and various initial cloud density (if Terrain from octree is used) need different filters settings. Both filtering methods are better to use with cloud created by Terrain from voxels due to its regularity.

Manual Adjustment

The above-described terrain extraction can be also adjusted by manual editing. Go to Terrain → Manual adjustment and select terrain cloud for editing. Set a name for a new cloud where points removed from the terrain cloud will be saved. Pressing the "x" button activates the selection box. The left mouse button makes the selection. The designated points will be removed from the terrain file and stored in the separate file defined above. For a step back press the undo icon → in the top panel. There is also a split function that creates user-defined strips of terrain for easier point selection. Arrows switches among the strips. The strip selection is finished by clicking at the split function again. Stop EDIT icon saves all changes in the original (input) terrain cloud.

IDW Interpolation

Areas with missing terrain points (typically shaded by thick trees or located right under the scanner during scanning) may be filled in by Inverse Distance Weighted interpolation (IDW) from the surrounding terrain points. Go to Terrain → IDW to select an input terrain cloud for interpolation. Set a resolution of interpolation in cm, a number of the closest points (n) to be included in the interpolation. Name the new interpolated terrain cloud that will be free of empty areas. The interpolation uses n-closest points of the original terrain to estimate the Z value of the new terrain point.

Slope

The slope analysis is a method that evaluates the terrain slope. This function is reachable via Terrain → Slope. The terrain slope, in general, is defined as the first derivative of the terrain. The slope is thus calculated from the height difference based on a user-defined sector. Users can choose between the sector defined by the nearest neighbors or the radius. Each point of the selected terrain cloud is then compared to the others within the defined sector and the mean slope value is computed. The result is stored as a field of intensity in degrees or in percent, where 45 deg represents 100%.

Aspect

The aspect is a method for evaluation of the slope orientation for a given terrain layer. There is defined surrounding for each point of the terrain layer where the average orientation is calculated. The result is given in degrees, i.e. -180 to -135 = West, -135 to -45 = South -45 to 45 = East 45 to 135 = North 135 to 180 = West. The result is stored as a field of intensity.

Hillshade

This function returns shaded relief of a given layer of terrain. The values of the resulting cloud range from 0-255 according to the azimuth and orientation.

Terrain Features

This is a method that looks up identical places in the field based on their parameters. Identical places are defined by their value in the field of intensity, the number of points, the size of the major axes, the axes ratio, the size of the maximal area (the concave), the convex, and the convex/concave ratio. The result is not saved to disk nor project as there is the Export Features function to do it.

Feature Table

The feature table displays a table of attributes for each terrain feature.

Export Features

This function exports the terrain feature information. The user selects the directory to save the file, delimiters of fields, and files to export. There is an option to choose between an attribute file and Convex/concave polygon. The attribute file is a file with the position of the centroid of selected voxels and its attributes. Convex/concave polygons are polygons delimiting voxels of given features. These are saved in the VKT format.