Grouping Routines
The aim of grouping is to identify points related with each other. For example, identifying a group of points representing the same object, like a single tree or a car. In research, grouping is often called segmentation. The grouping is done based on different methods, such as plane fitting, watershed algorithm, or 3D spacing between points.
Many of the group assignment routines rely on the distance from ground value and normal vector information. Both must be computed for the points before the command is started. Use Compute distance command for loaded points or Compute distance macro action with setting Distance: To ground class in order to compute the distance value. Use Compute normal vectors command for loaded points or Compute normal vectors macro action in order to compute normal vector directions.
Usually grouping all objects in a point cloud requires running several grouping routines. Most typical is to 1. group Planar surfaces, 2. group By tree logic, and 3. group By density, with the first action clearing existing grouping, and the following actions keeping the previous results, and creating new groups for remaining points.
The group number can be used for the visualization of points and for classifying points. There are various automatic classification routines that rely on grouping. The tools from the Groups toolbox can be used to manipulate the groups manually.
The group number can be stored in TerraScan FastBinary files. Switch on the Group attribute in the Save points as dialog or in the Attributes to store dialog of a TerraScan project in order to store the group number for project block files.
In addition, there are some classification routines which can be used in macros only. This Chapter describes the basic logic used in the classification routines.
In TerraScan Lite, only tools for modifying groups are available.
The Assign groups sub-menu in the Group pulldown menu of the TerraScan window offers a number of grouping routines that add group information to points.
•Assign groups
