Automatic classification mode allows classification of points using special algorithms which extracts from the dataset points representing a given land cover class. The final result of the automatic classification depends on a number of parameters, such as point cloud density, type of terrain, the sequence of subsequent classification steps, or building dimensions, for example an algorithm that produces great results, e.g. in flat areas, can incorrectly classify points in mountainous terrain, etc. In order to obtain the best results, the parameters of particular algorithms should be adjusted for a given data set.
Classification steps
Point cloud classification is carried out in three main steps:
3.Classification of vegetation
The figure below illustrates the process of classification for the exemplary point cloud:
The classification process has to run exactly in the steps specified above. There is no way to extract buildings without having properly extracted ground points. Similarly, vegetation cannot be extracted correctly unless ground and building points are not classified. Understanding the above-mentioned rule is crucial for correct interpretation of classification results.
For example, one can try to solve the failure in the classification of vegetation by changing vegetation classification parameters. However, errors in vegetation extraction may be related to the improper identification of ground points. Consequently solving the problem requires re-classification of ground (and buildings) with different settings.
In some applications, it is required to extract not all point classes. For example building Digital Elevation Model (DEM) requires only ground points classification. To see how to avoid extraction of unnecessary classes visit description of General Settings.
Class naming convention
Note: In the current version of the software, the class numbers are fixed according to the ASPRS LAS file standard. Class numbering is specified in the table below and can't be changed by the user.
ID |
Content |
0 |
never classified |
1 |
unclassified |
2 |
ground |
3 |
low vegetation |
4 |
medium vegetation |
5 |
high vegetation |
6 |
buildings |
12 |
overlap points |
Datasets with already classified points
The general rule is as follows: in the initial stage of classification, class parameter for all points in the dataset are reset, except points having class id = 12 (overlap points). Overlap points will never be subjected to classification process and will preserve their class id (=12) through whole classification procedure. This rule does not apply if You specify output classes set which do not involve ground points (see table in the section: General Settings: no.: 3,4,7).
Hotkey: Edit.SettingsDialog