Relative Accuracy
Relative accuracy, also known as “swath-to-swath accuracy” or “interswath consistency” is the measurement of how well overlapping areas of data collection match each other. For LiDAR datasets, although offsets may occur in any direction, we measure this in the vertical only. Relative accuracy is influenced by many factors, many of which can be modeled and corrected for. Relative accuracy statistics are highly dependent on mission specific properties, such as trajectory quality, boresight quality and software/workflow employed, and are typically an indicator of those factors.
These factors have a significant impact on relative accuracy:
Systemic Calibration (Lever Arms & Boresight)
Trajectory Quality
Processing software and workflow
Measuring Relative Accuracy
Ground surfaces are computed on a per-flightline level using a PTD algorithm in SpatialExplorer.
Using ground classified points serves to eliminate above ground objects and other unreliable areas of the point cloud.
Generally this classification is performed across the entire dataset, rather than user defined regions, to reduce complexity in the routine.
Default values are used for the PTD algorithm.
Then each surface is compared to all ground classified points from all overlapping flightlines.
Elevation values are interpolated from a TIN surface in order to perform the comparison. The differences are recorded for each flightline surface.
These differences are summarized by an average signed delta-z and an average unsigned magnitude per flightline.
Finally a project-wide average magnitude is stated to characterize the entire dataset.
Industry References
ASPRS outlines a few options for how to measure swath-to-swath accuracy.
Elevation values for each flightline can be compared to surveyed check points.
Elevations for each flightline should be assessed using a TIN of it’s surface points, interpolating an elevation value at each checkpoint X,Y location and recording the residual difference.
These differences can be summarized to produce an RMSDz value, where results show relative accuracy between flightlines based on their differences from common checkpoints.
Surface raster’s can be compared to achieve a statistical understanding of how flightline elevations differ throughout a scanned area.
Elevation raster or surfaces per flightline should be created from single return hard surfaces areas.
Delta-Z rasters are then computed by subtracting one flightlines surface from another.
The cell values from the Delta-Z rasters are then used to compute RMSDz
In either case, ASPRS suggests that relative accuracy be assessed in all areas of overlap, using non-vegetation regions only, and ignoring zones of quickly changing terrain or vertical artifacts.
ASPRS Positional Accuracy Standards for Digital Geospatial Data, 2014. http://www.asprs.org/wp-content/uploads/2015/01/ASPRS_Positional_Accuracy_Standards_Edition1_Version100_November2014.pdf
The USGS Guidelines are very similar to ASPRS in this case as well. Similar to their precision methodology, USGS lays out a raster based methodology where:
Difference rasters should be produced for regions of overlap where terrain slope is less than 10 degrees.
The cell size for analysis should be average nominal point spacing(rounded up to an integer) x 2
Differences rasters should be signed
Difference rasters should be statistically summarized to show:
RMSDz
Min difference
Max difference
LiDAR Base Specifications, 2021. rev A. https://www.usgs.gov/core-science-systems/ngp/ss/lidar-base-specification-data-processing-and-handling-requirements#positional-accuracy-validation
Last updated