Point Cloud Segmentation

Segmentation is an automatic tool that breaks the point cloud into smaller groups calledsegments based on the spatial and attribute relationships between points in the point cloud. By segmenting the point cloud(s) into smaller sections and applying a segment ID, users are better able to select points by segment for manual classification or further editing. The goal of this tool is to provide users the ability to separate objects in the point cloud apart from the neighboring points to assist with classification. Objects, in this context, can be buildings, ground, cars, road paint, transmission towers, things that have distinct attributes or structure from their neighbors and can be detected by the tool. Objects are delineated based on the attributes and thresholds set in the tool by the user.

This tool requires Global Mapper Pro.

After using the Point Cloud Segmentation tool, the Lidar Draw Mode setting will be changed to Segment ID to show the point cloud colored by segment. Colors are assigned at random. Unlike the automatic classification tools, the segmentation dialog will remain open after the process has been completed. Running this tool again on the same or a different point cloud during the same segmentation session will continue the counter of segment IDs so there are no repeated values.

Tiling data processing: When working with large data sets, Global Mapper will work to manage memory and machine resources. Internal tiling of the data set may be performed if the program estimates insufficient memory. If internal tiling is needed for resource management, the user will be prompted when the process is executed. After processing, the Segment ID settings can be used to merge segments.

Attributes are on the left side of the dialog, and thresholds, called Segmentation, are on the right.

Resolution

The resolution at which to analyze the point cloud. This distance, specified in linear units or a multiple of the layer point spacing, will determine the local neighborhood size used for any given point when completing this analysis. Put simply, neighborhood size determines how far from each point the segmentation tool should look when comparing the point to its neighbors. If the neighborhood is too small, the segmentation tool might not detect the full structure of the object. Too large, and smaller structures and details may be glossed over for a less accurate result.

Attribute Evaluation

Below are the attributes that can be used to evaluate the point to point similarity that informs spectral partitioning. Select which attributes to use by checking the box next to attribute and assign a weight for the attribute. A higher value entered will result in that particular attribute being more heavily considered in the point to point similarity measure. Attributes of a point are based on statistics of a local area (neighborhood) with extent defined by the resolution.

Position refers to the X/Y/Z or Lat/Lon/Elevation position of each point return. This option is always enabled as the positional relationship between points will always be considered.

Normal considers the direction perpendicular to the surface the point is representing. Similar normal values over a local area indicate the points are representing a consistently shaped feature.

The normal of a lidar point can be depicted as a ray or arrow perpendicular to the surface the point represents. Looking at the cross section view of a roof, the plane of the roof is the surface represented by the points, and the normal for a single point is shown perpendicular to this plane moving away from the point.

Intensity is the strength of the return collected. Most lidar systems record the intensity of the returning pulse. Similar materials will have similar intensity values making it useful when looking for groups of like points within a layer. Use the Lidar Draw Mode setting to visualize intensity values on a black and white gradient in the point cloud.

Return Number looks for similarity in the return number of each point return in the data set. Return numbers typically range from 1 to 5. Solid, uncovered surfaces typically have larger numbers of first and only returns while vegetation and layered objects have a larger mix of return numbers. Use the Lidar Draw Mode setting to visualize returns in the point cloud.

Curvature analyzes the curves created by the points in a local neighborhood. Consistency in curvature values will help indicate that points likely belong to the same object.

Color considers the RGB color attribute associated with returns in the point cloud. In some data sets, similar colors may indicate that points represents parts of the same feature and should be included in the same segment.

Segmentation

Connectivity Threshold is the threshold value for algebraic connectivity that is used to determine where to cut when dividing segments. A larger value will result in more segments. In general, a low value (such as a decimal below 1) would be used for large data sets where many different types of points are included. A high value (such as 50) would be used when trying to segment within a category or class, such as further segmenting the ground.

Minimum Number of Point in Segment determines the minimum number of points needed in a cluster for a segment ID to be assigned. This value should be set with the point spacing of the data set in mind. A smaller number will result in many more smaller segments, while a larger number will result in fewer, larger segments.

Maximum Number of Standard Deviations narrows the range of point associations to be considered. A larger number of standard deviations will include more weakly connected points in the same segment. A lower value will likely create fewer segments but the points in each will be more closely related.

Maximum Curvature determines the maximum allowed curvature within a neighborhood in order for a cohesive segment to be identified.

Curvature is determined by analyzing how the position of adjacent points relate to one another. The maximum curvature threshold dictates the limit of curvature in a local area for a point to be considered part of the same segment.

Looking at the cross profile of a car below, the point representing this one vehicle are split into two segments. The break between these segments happens at the top of the windshield area where the curve along the surface of the car is at its sharpest over a local area. If the curvature value were to be increased for this data, the car could be assigned a single segment ID.

 

Segment ID(s) Settings

The Segment ID(s) Settings button opens a dialog that allows users to select existing segments to exclude from the next Segmentation analysis, Merge segments, or Rename segments.

Note: Closing the Segmentation tool will reset the numbering scheme. When the tool is reopened and another analysis performed, the new segment ID numbers will not start where the last set left off, instead it will start over and overwrite the existing segment ID values. To keep your segment ID values, leave the Segmentation tool open.

Particular phenomenon or objects can be excluded from a subsequent segmentation process by selecting them from the list, for example a segmentation looking to target vehicles in a parking lot might explicitly exclude some other signs or structures in the parking lot. The tool can also be used to merge segments together, for example, merging two or more segments that represent a car into one segment for the entire car. Use the Select Lidar Segments Tool (link) to select them on the map, then check the Use Only Segments Selected in Digitizer to sync with this editing tool.

Available Segments

The Available Segments section lists all currently identified segments in the point cloud layer(s). Use the check boxes to select segments to merge or exclude. Selected segments will populate the Selected Segments section of the dialog.

The option to Use Only Segments Selected in Digitizer will hide all segments in the Available Segments list and populate the Selected Segments list based on the digitizer selection.

Double click on a Segment ID value in the Available Segments list to alter the segment name.

Exclude or Merge

Exclude Selected Segments from Segmentation Process

To Exclude Selected Segments from Segmentation Process select this option and leave the Segment ID(s) Settings dialog open as you execute the segmentation analysis. The selected segments listed in the Selected Segments section will be ignored in the segmentation process and remain unedited.

Merge Selected Segments into New or Existing Segment

To merge segments, select the segments to merge through the Available Segments section, and enable the Merge Selected Segments into New or Existing Segment option under Exclude or Merge.

When this option is enabled theSelected Segments list will show the selected segments along with a new segment ID containing zero points. In the Selected Segments section enable the check box for the segment ID you would like to keep and click Merge Selected Segments to assign all the points in the Selected Segments to the checked segment ID to keep.

Selected Segments

The Selected Segments section of the dialog lists the segments selected to either merge or exclude.