Point clouds, with their rich, three-dimensional data, have revolutionized how we map and analyze the world. Yet, the true potential of point clouds is often unlocked through their meticulous classification. Enter Global Mapper Pro, robust GIS software that not only manages but also excels in enhancing the capabilities of point cloud data through advanced sub-classification techniques.
The article below explores how Blue Marble Geographic’s recent advancements in point cloud sub-classification offer a transformative approach for professionals seeking to extract the maximum value from their data. By breaking down complex datasets into more granular, easily manageable segments, this powerful tool allows users to dive deeper into their analysis, providing clearer insights and facilitating more accurate decision-making.
The ability to create custom point cloud classifications opens the door to another level of point cloud management: sub-classifications.
Creating Custom Feature Classifications in a point cloud is a cutting-edge new tool originally released in version 25 of Global Mapper Pro. This new functionality allows users to leverage machine learning techniques to train a custom model. This model looks for specific features in a point cloud for classification. Assigned classifications can be any built-in classes such as ground or building, or a new class such as sidewalk or fire hydrant. What happens though, when a class needs to be further refined, or have other classes identified within it? That is where sub-classifications become beneficial.
What is a Sub-Classification?
Point cloud sub-classifications, released earlier this year in Global Mapper Pro v25.1, allow users to refine an existing class. Sub-classes are part of an already existing class. Once trained as a custom feature classification, the sub-classes are defined with their own classification code, name, and color – uniquely identifying the feature of interest.
In this point cloud of a golf course, subclassifications of Ground points include cart paths (gray) and golf greens (sea green).
Sub-classifications can be applied to a variety of use cases, from identifying cart paths on a golf course, to particular roof structures in buildings. In the below example, we’ll take a look at an airport and attempt to identify runway paint markings on the runway, which is already classified as ground.
Creating a Sub-Classification
When creating a custom classification, whether an initial class or a sub-class, the process focuses on using attributes of the training samples to develop what’s called a signature. These attributes are similar to what the segmentation tool uses to identify clusters of points: physical structure, return number, etc. Once created, this signature is used to identify subsequent points which fit that criteria. The point cloud below has RGB values, which will be a key attribute in distinguishing the paint from the runway. This process can also rely on segmentation outputs to help build signatures. For more information, see: How to Train a Custom Point Cloud Classification in Global Mapper Pro.
The goal of this workflow is to classify the painted lines as a subclassification of the existing Ground class.
The first step in the process was to run the built-in automatic classifiers. This allowed for the identification of ground and vegetation, along with buildings and powerlines (not pictured). You’ll notice that the ground classification encompasses runway, grassy areas, and other similar flat, ground-level features.
Use the automatic classification tools first to exclude any irrelevant points from processing, such as the vegetation pictured here.
In many scenarios, the above classification is exactly what is intended. Ground is classified regardless of whether or not that ground is paved, grassy, or otherwise. In situations where more detail is required, sub-classification is the next step.
A sub-classification process is set up nearly identically to a custom-trained classification. The major difference is that the setting for the option “Is subclass of” will be set to ground. Once set, the classification can be trained by selecting points that represent the feature within the ground points to be classified. In this example, when collecting training samples, displaying the data by RGB allowed for clearer visualization of the paint markings. Segmenting the point cloud ahead of time can help with creating clusters of points for selection.
The sub-classification is trained based on the signature of selected points in the point cloud.
Once the training is complete, the classifier can be run again, this time focusing on the newly created runway markings sub-class which was created previously. The result of which is the independent classifications of the paint markings. The white color was randomly assigned and can be customized by the user as needed.
Tip: Classification color can be edited from the Lidar section of the Configuration menu.
The white paint lines are now a subclassification of the ground. They are functionally the same class but still distinguishable in visualization and analysis.
Once assigned a classification, these sub-class points can be distinguished from the parent class without being excluded from it. For example, if we were to create a digital terrain model using only ground points, because the paint points are a subclass of ground, they would be included in the model creation as well. Sub-classifications give you more control over your data for fine-tuning and more accurately analyzing the output.
Try Global Mapper Pro with a free 14-day trial today.
