The LIS LiDAR Point Cloud Classification Plugin provides methods to classify airborne LiDAR collections to ground, building, and low/medium/high vegetation.
- Ground Classifier:
The Ground Classifier follows a progressive TIN (Triangulated Irregular Network) densification approach. In a predefined raster, seed points are determined to form an initial TIN. According to the classifier parameters the TIN is then iteratively refined by individual points of the point cloud, until no point matching the insertion criteria is found. All points within a certain distance to the TIN are then labelled as ground. All other points are labelled as unclassified.
- Building and Vegetation Classifier:
Building points are confined to planar surfaces found within the point cloud. Therefore a robust plane fitting algorithm is applied to all unclassified points of the point cloud to detect surfaces of planar objects. Surfaces with similar orientation values are then grouped to objects by region growing methods. Roof installation and points below the roof points within a certain distance can then be relabelled as roof details and facade points.
Vegetation is sliced into three vegetation heights (low, medium, high). Based on a 3D search radius and the point count within this radius, non-vegetation points can be labelled as unclassified.