Classification algorithms
- Jonathan Duguay (Unlicensed)
Classifying point clouds to predefined classes is often one of the most important processes to be done with lidar or point cloud data. The points in a project can be classified to 128 different classes, which allows the user to segment, isolate or filter targeted data and help visualization and analysis. The data can be classified manually by selecting points with the selection tool and classifying them or the classification can be done automatically using algorithms.
Many automatic classification algorithms are available with VisionLidar to speed up classification. Even though the classification algorithms are automatic, they each have a series of parameters to set, since every project can have types of terrain or may come from a different scanner. Sometimes, tests are needed to find the ideal combination of parameters for a specific project. It’s therefore advisable to limit the detection to a small area before applying the detection to the whole project. This functionality is included in all the classification algorithms except when computing normals.
For best results, it’s recommended to classify points in the following order:
There are also two other process that allows a classification of several points at once:
If surfaces of ground or buildings were created, it's possible to classify points around those Surface objects
If clusters were detected, it's possible to select several clusters with a filter and classify them in a selected classe