Improving the Quality of Lidar Data in Global Mapper
As with any commodity or raw material, the quality of lidar or point cloud data has a direct bearing on the quality of its derivative products. A point cloud in which the positional accuracy is below the required standard will produce an inaccurate Digital Terrain Module (DTM); misclassified points will result in a misrepresentation of any extracted vector features; an overly dense point cloud will create unnecessary coarseness in the corresponding raster surface; noise points falling either above or below the expected elevation range will likely corrupt any subsequent analysis or data processing workflow.
Global Mapper’s Lidar Module, which is available as an optional add-on to the software, provides a plethora of tools for editing, filtering, and generally improving the quality of point cloud data. The updated point cloud can subsequently be utilized in the Module’s extensive range of analysis tools, or it can be exported to create an improved version of the lidar data as a final deliverable.
Filtering During Import
Anyone who has imported a point cloud into Global Mapper will have seen that the simple act of loading the data triggers a dialog box in which numerous filtering options are available to limit the loaded point cloud to a more relevant subset. While this is a convenient utility for filtering while loading, in most situations, a preliminary analysis of the characteristics of the full dataset is required to make informed decisions about which points to remove.
The Lidar Import dialog box provides numerous options for filtering the data before it is rendered in the map view
In many cases, the most important analysis is visual. Does the point cloud cover the extent of the area of interest? Are there any significant coverage gaps? When displayed in the 3D Viewer, are there any obvious vertical anomalies? A more methodical approach to this preliminary quality control process involves scrutiny of the layer’s metadata, which is accessed from the Control Center. The metadata reveals a wide variety of information about the layer, including the elevation range, average density, point spacing, coverage area, collection date, and perhaps most importantly, the classifications to which the points have been assigned.
Armed with this information, it is now possible to embark on a series of data improvement procedures to address any perceived deficiencies in the point cloud.
Broadly speaking, the process of improving point cloud data falls under two categories: editing and filtering. Editing typically involves updating the characteristics of individual points or, as is more likely, a selected group of points, while filtering infers the removal of points that are not needed or that are deemed to be erroneous. Often, these two procedures are applied in tandem.
Editing Point Classifications
The most common editing procedure applied to point cloud data is applying or updating the classification values to reflect the type of surface represented. The standard classifications available within the lidar format specification include ground, building, low, medium, and high vegetation, and water, along with numerous others. In some point clouds, especially those created from photogrammetric analysis, all of the points will be initially unclassified, and in order to perform any type of systematic analysis, relevant classifications will need to be applied.
Using the Lidar Module, point classifications can be updated manually by selecting the required points and applying the updated classifications using one of the preconfigured classification buttons in the toolbar. While this approach might be considered impractical for a wide-ranging reclassification project, it is useful for quickly updating a limited number of selected points, especially when the points are rendered in the Path Profile, or cross-sectional view.
Points representing the roof of a building selected and manually reclassified in the Path Profile display
A more efficient alternative to manual reclassification is to use one of the Module’s automatic classification tools, which geometrically analyze the point cloud for the purpose of identifying and reclassifying certain types of points. Specific algorithms can be applied to detect points representing ground or bare earth, high vegetation or trees, buildings, above-ground utility cables, and power poles. These reclassified points can subsequently be used for DTM creation or feature extraction.
Points representing buildings and trees automatically identified and reclassified
Editing the X, Y, and Z Values
Another characteristic of the point cloud that can be edited or updated is the three-dimensional position or location of the points. Typically, a lidar file will not need to be shifted, however photogrammetrically derived point clouds are more likely to require positional adjustments. In the event a point cloud is not precisely aligned with an underlying reference layer or with another point cloud, the Lidar Module offers several tools to improve positional accuracy:
In each of these procedures, the coordinates and/or the elevation assigned to each point will be updated to reflect the shift.
A photogrammetrically generated point cloud positionally adjusted using the Fit Lidar tool to align with an existing lidar layer
Almost without exception, a point cloud will need to be filtered before beginning any analysis, surface generation, or feature extraction workflow. The filtering process can be based on the geographic extent of the data or, more commonly, on its inherent attributes or characteristics.
As with any vector data, a point cloud can be cropped to the extent of a preselected polygon providing a simple method to remove redundancy in any subsequent processes. This is perhaps the most useful of the filtering options available during the data import process but it can also be applied after the data has been loaded.
A lidar point cloud filtered or cropped to the extent of a project area
Point proximity can also be used as the basis for filtering. The Lidar Module’s thinning tool removes points based on a defined density or spacing requirement, while retaining just the point with the maximum, minimum, or average elevation within the immediate neighborhood. This is especially useful when working with high-density, photogrammetrically created point clouds.
Noise points are defined as those that fall outside the normal elevation range within a specified local area. Noise points may be present because of issues with the data collection process, physical anomalies in the target area, such as birds flying below the aircraft, or for numerous other reasons. The Lidar Module’s noise identification tool can either reclassify noise points, or it can immediately remove or delete them, based on a prescribed set of parameters and elevation thresholds.
Noise points falling above and below the expected local elevation thresholds reclassified and displayed in the 3D View
Filtering by Classification
The first step in the process of transforming a point cloud into a raster DTM is, by necessity, the elimination of points that are not classified as ground. Retaining non-ground or unclassified points will create a surface that includes above-ground features and will not reflect the bare-earth morphology of the target area. Filtering in this context is a simple class selection process, which can be applied to either the on-screen display of the points, or during the gridding or surface creation process. In either case, the elevations assigned to the pixels in the resulting raster layer will be derived from the selected subset of the point cloud.
Non-ground points removed prior to the creation of a Digital Terrain Model
Point Cloud Exporting
While Global Mapper’s Lidar Module offers a wide range of editing and filtering tools, it is important to note that the improvement procedures described in this article do not automatically apply to the original files or loaded layers. In order to save the results of any of the point cloud improvement procedures to a file, the data must be exported in the appropriate point cloud format. The new layer will inherit all of the modified point characteristics and will be limited to those points that have been retained after filtering or deleting points that are not required.
The Lidar Module’s powerful analysis, extraction, and surface generation tools understandably garner the most attention and accolades from geospatial professionals. However, it is important to acknowledge the vital role that the Module’s editing and filtering tools play in these procedures. The better the quality of the raw material, the better the final product.
If this blog piqued your interest and you’d like to learn more about Global Mapper and the Lidar Module, download a 14-day free trial and request a demo today! For more information, contact us.
If you enjoyed this blog, you may also find these other resources useful:
Improving the Quality of Lidar Data in Global Mapper