Use of this function requires a license for Whitebox Workflows for Python Professional (WbW-Pro). Please visit www.whiteboxgeo.com to purchase a license.
This tool provides a basic classification of a LiDAR point cloud into ground, building, and vegetation classes. The algorithm performs the classification based on point neighbourhood geometric properties, including planarity, linearity, and height above the ground. There is also a point segmentation involved in the classification process.
The user may specify the names of the input and output LiDAR files (input
and output
). Note that if the user does not specify the optional input/output LiDAR files, the tool will search for all valid LiDAR (.las, .laz, *.zlidar) files contained within the current working directory. This feature can be useful for processing a large number of LiDAR files in batch mode. When this batch mode is applied, the output file names will be the same as the input file names but with a '_classified' suffix added to the end.
The search distance (radius
), defining the radius of the neighbourhood window surrounding each point, must also be specified. If this parameter is set to a value that is too large, areas of high surface curvature on the ground surface will be left unclassed and smaller buildings, e.g. sheds, will not be identified. If the parameter is set too small, areas of low point density may provide unsatisfactory classification values. The larger this search distance is, the longer the algorithm will take to processs a data set. For many airborne LiDAR data sets, a value between 1.0 - 3.0 meters is likely appropriate.
The ground threshold parameter (grd_threshold
) determines how far above the tophat-transformed surface a point must be to be excluded from the ground surface. This parameter also determines the maximum distance a point can be from a plane or line model fit to a neighbourhood of points to be considered part of the model geometry. Similarly the off-terrain object threshold parameter (oto_threshold
) is used to determine how high above the ground surface a point must be to be considered either a vegetation or building point. The ground threshold must be smaller than the off-terrain object threshold. If you find that breaks-in-slope in areas of more complex ground topography are left unclassed (class = 1), this can be addressed by raising the ground threshold parameter.
The planarity and linearity thresholds (planarity_threshold
and linearity_threshold
) describe the minimum proportion (0-1) of neighbouring points that must be part of a fitted model before the point is considered to be planar or linear. Both of these properties are used by the algorithm in a variety of ways to determine final class values. Planar and linear models are fit using a RANSAC-like algorithm, with the main user-specified parameter of the number of iterations (iterations
). The larger the number of iterations the greater the processing time will be.
The facade threshold (facade_threshold
) is the last user-specified parameter, and determines the maximum horizontal distance that a point beneath a rooftop edge point may be to be considered part of the building facade (i.e. walls). The default value is 0.5 m, although this value will depend on a number of factors, such as whether or not the building has balconies.
The algorithm generally does very well to identify deciduous (broad-leaf) trees but can at times struggle with incorrectly classifying dense coniferous (needle-leaf) trees as buildings. When this is the case, you may counter this tendency by lowering the planarity threshold parameter value. Similarly, the algorithm will generally leave overhead power lines as unclassified (class = 1), howevever, if you find that the algorithm misclassifies most such points as high vegetation (class = 5), this can be countered by lowering the linearity threshold value.
Note that if the input file already contains class data, these data will be overwritten in the output file.
colourize_based_on_class, filter_lidar, modify_lidar, sort_lidar, split_lidar