Built Terrain feature extraction in Point Clouds
This preliminary investigation explores the classification of point clouds obtained from UAV images and created through photogrammetry. In addition to considering red, green, and blue color data, the study emphasizes neighborhood search methods. It investigates the application of computed geometric feature attributes to distinguish between natural terrain features and human-built features.
Normal Direction
The orientation of each point to its neighboring points allows to compute normal vectors for each point. These vectors can be decomposed into X-, Y- and Z components and can be used to evaluate slope conditions or determine if a point has a rather vertical neighborhood or planar neighborhood to surrounding points within a defined search radius.
Planar and Linear Features
“Planarity” and “Linearity” are geometric feature attributes that mostly relate to human agency, rather than they do to natural form and terrain. Although these features can also be detected in natural formations, these attributes will reflect the highest values in built environments and objects such as houses, powerlines, etc.
Detecting Geometric Features
