The classification is a process when the groups of LAS points receive a class value; based on the class, the extraction and visualisation can be performed in a more accessible way as the user can focus on those groups of points which are relevant for the current extraction task. The classification is most commonly performed during post-processing or with other 3rd party software using automated solutions. The PCS software offers manual or semi-automatic classification tools. The PCS software is not intended to perform automated classifications such as ground classification, tree classification, rooftop classification, etc. The main goal of the classification tools inside PCS is to help the users fine-tune already classified datasets or support the vector extraction with noise filter or ground fine-tune tools. PCS is also suitable for annotating training datasets from point clouds for machine learning scenarios.
The LAS files store the classification value in the class attribute. Each point inside the LAS file has a class value, which can be between 0 and 255. These classes can have a name and colour inside the PCS (these are not stored inside the LAS; they are unique for the PCS). To inspect the class names and colours, inspect the class filter window.
For an easier understanding, classes are similar to CAD layers: you can turn on or off each class (inside PCS, separately for 2D and 3D windows) using the class filter window. Some tools can also consider the classes for running a tool (export, information, density, extraction, TIN, etc.).
The classification trends are different for each manufacturer or sensor type. TLS and SLAM data are commonly not classified (which means all points are located in class 0 or 1), MLS data is frequently classified minimum for a ground class and others, and ALS data is commonly classified for ground, vegetation and roofs. These automated classifications cannot always be trusted; crosscheck the classification quality before blindly trusting in the ground class.
The classification can be performed inside PCS only if all points are loaded into the memory upon loading the LAS files into the software. Read the respective article for more info about All points in memory. If no point cloud loaded into the project with all points in-memory options, the classification toolbars are remain unavailable. If there is a point cloud with all points in memory but the classification toolbar remains unavailable, set the LAS file active again.
The software can visualise the point cloud based on class values; see the point cloud toolbar article for a description. Also, read the class filter window to understand how to control the class visualisation.
Please note that the classification actions can be undone (but not redone). The software has a designated memory space to store classification actions. If this storage gets full, the software will prompt the user to clear the storage in the memory. When this happens, the previous classifications cannot be undone. The software automatically saves the classification to the LAS files, but some functions might not have this feature, so it is strongly recommended that the LAS files be saved upon modification. The Save all function will always prompt the LAS files which have not been saved. To get information about the currently available buffer size, check the Main Classification toolbar - Undo More tools, as the last classification operations and the available buffer size are visible there. Additional buffer memory can be appended if more undo is required during the operations.
Working with classes commonly results in the users “losing the point clouds”.a point cloud has It is a common support issue that the users load point clouds, but they are not visible in 2D and/or 3D view. This is commonly a classification-setting issue. Please read the respective debug article.