July 6, 2022

Point Cloud Segmentation for Road Identification

Written by: Mackenzie Mills


Numerous tools for lidar editing, filtering, and classification can be found within the advanced geospatial program, Global Mapper Pro. Making the analysis and processing of point cloud data easier, Global Mapper Pro includes tools for the automatic classification of points representing ground, buildings, trees, poles, and powerlines. These tools work by adjusting a set algorithm designed to identify expected shapes by user-input parameters. Of course, points can also be manually classified using the ASPRS-defined lidar classes or custom classes that reflect the unique characteristics and features found in a specific point cloud. To assist in the process of manual classification, Global Mapper Pro’s Point Cloud Segmentation by Spectral Graph Partitioning tool can identify clusters of points representing features outside of those available through the automatic classification tools.

Allowing for increased customization in feature identification and selection for classification, Global Mapper Pro’s Point Cloud Segmentation tool considers a variety of point attributes the user chooses, and based on a calculated similarity, identifies contiguous clusters of points defined by user-input thresholds. While the result of the Point Cloud Segmentation process is not a classification applied to points, instead the returns receive a Segment ID value based on the analysis. The applied ID value makes the identified features easy to select with the Select Segment(s) tool for subsequent manual classification. While Segmentation can be used to identify ground, buildings, and other standard features, it can also be used for more customized feature identification, such as cars or dirt paths.

Road Identification

While paved roads are distinct in color and texture from the surrounding ground, the points representing roads in the point cloud are ultimately considered ground by Global Mapper Pro’s automatic classification. The method for ground classification pays more attention to the shape the points create by analyzing local changes in elevation. To create sub-classifications for different types of ground cover, changes in additional point attributes need to be considered.

With the ground already classified in this point cloud, a filter is applied to remove the non-ground points. Now displaying only the Ground classified points, the visualization is changed to reflect different attribute values by selecting alternate Lidar Draw Modes. When viewed by intensity, the roads become clearly visible since they are all paved with asphalt. Because the asphalt is a relatively dark color, more of the laser’s pulse is absorbed resulting in a relatively low-intensity return.

Viewing the filtered point cloud by intensity clearly shows the roads.

In the Segmentation tool, each point’s position, intensity, and curvature will be considered for this process. Since the intensity values for the road points are distinct from the surrounding returns, the intensity is the highest weighted attribute in this analysis and because roads are a subclass of ground, the connectivity between identified points should be stronger. With the roads being a connected network throughout the extent of the point cloud, the process will only identify segments with a larger minimum point count.

The Segmentation tool allows users more control over what attributes are considered in the analysis.

With the Point Cloud Segmentation executed in Global Mapper Pro, the Lidar Draw Mode automatically changes to Color by Segment to clearly show the identified clusters of points. The road segments are selected with the Select Lidar Segment(s) tool, which selects an entire segment of points with a single click. Multiple selected segments are then merged through the Segment ID Settings option within the Point Cloud Segmentation dialog box.

Using the Select Lidar Segment(s) tool, identified segments are selected and merged.

Finally, the road points identified and selected as a segment, are easily classified using the option to Edit Lidar Points, or with the Change Lidar Class tool. The color assigned to the lidar Road class is also customized through the Filter Lidar Data tool, in order to show the identified roads in a bolder color when the full point cloud is colored by classification.

Shown in 3D by classification, the classified road points are clearly seen in the point cloud.

Global Mapper Pro’s Point Cloud Segmentation by Spectral Graph Partitioning tool can be used in many contexts to identify features represented in a point cloud, but not explicitly covered in the program’s automatic classification tool. Using Segmentation in conjunction with the visualization options and manual editing capabilities of Global Mapper Pro provides an efficient workflow for custom feature identification.

If you are interested in exploring Point Cloud Segmentation in Global Mapper Pro, download a 14-day free trial today! If you have any questions, please contact us.

If you enjoyed this blog, you may also find these other resources useful:

Point Cloud Segmentation Analysis in Global Mapper

Ask the Experts: How do I identify and classify custom features from a point cloud in Global Mapper?

Companies using Blue Marble’s geospatial technology