LiAirV: Data Processing
LiAir V data processing operations must be carried out in GVI’s LiGeoreference software. LiGeoreference provides a comprehensive software solution for the accurate georeferencing of LiDAR data collected from GVI’s mobile, UAV, and airborne 3D laser scanning systems. LiGeoreference integrates the raw range measurements of the Livox LiDAR sensor with the LiAir V’s GNSS/INS derived position and orientation estimates to produce 3D point clouds mapped to a user-defined coordinate system. The software also allows for the fusion of 2D imagery with 3D point clouds and the generation of photo-realistic texturized representations of remotely sensed objects and areas.
GVI’s LiDAR360 software also is used in the LiAir V data processing workflows described in this Product Article. This point-cloud post processing program compliments the functions found in LiGeoreference. While LiAir V users are not required to purchase LiDAR360, it is highly recommended that as it will stream line the creation of specific high-value deliverables from UAV-LiDAR point clouds.
POS Proccessing
The first step in the LiAir V data processing workflow involves opening the .live file created for a given project and downloaded from the system to a computer with the LiGeoreference installed on it. The user will then have to define the Project Settings before running the POS processing module (LiNav) and using its outputs to snap the LiVox laser scan data to a target global coordinate system. Currently, LiGeoreference requires users to reference the global ellipsoidal World Geodetic Survey of 1984 (WGS84) datum and coordinate system when processing their raw POS data. This means users MUST have Base Station setup location data (latitude, longitude, and elevation) available to them in coordinates that also reference WGS84 datum and ellipsoid for the same epoch during which the base station data was collected.
Even though WGS84 is the reference system used by the Global Positioning System (GPS), many surveyors working in the United States will report base station setup locations that reference NAD83 horizontal coordinates and NAVD88 orthometric heights. Do not use these values as they will lead to errors in the absolute accuracy of the LiGeoreference processed results. North American users should also be aware that the popular Online Positioning User Service (OPUS) run by NOAA’s National Geodetic Survey does not report solutions in reference to the WGS84 earth-centered cartesian coordinates and ellipsoidal heights thus they must be transformed before they can serve to populate the Base Station Data fields of the LiGeoreference Project Settings. Current geodetic realizations of the International Terrestrial Reference System (ITRS) maintained by the IERS are still only metre-level consistent with WGS 84 and when we are talking about a system with +/- 5 cm accuracy capabilities, this magnitude of difference can be significant enough to introduce error into the LiGeoreference outputs and lead to incorrect data quality assessments.
LiNav provides a system performance report for the operator to examine the quality of the processed POS information. The html-formatted report contains plots that describe the availability of GNSS satellites, the status (fixed or float) of ambiguities, PDOP, VDOP and HDOP values, estimated accuracies of attitude, position, and velocity, as well as height and velocity profiles throughout LiAir V data collection.
Reprojections
By default, coordinates of the output point cloud generated by LiGeoreference will be in the WGS84/UTM (local zone) horizontal coordinate system and vertical coordinates will reference the WGS84 ellipsoid. Each LiDAR measurement will be assigned a Universal Transverse Mercator (UTM) Grid X (Easting) and Grid Y (Northing) value. Z-values (elevations) will simply be the WGS84 ellipsoidal height values in meters. The point cloud spatial reference system information will be stored in the header of the LiDATA or LAS/LAZ file exported from LiGeoreference.
It is possible for users to transform the default 3D coordinates of entire point clouds into a different coordinate reference system (CRS) using functions found under the Target Coordinate System tab of in the LiGeoreference Project Settings. The operations parameterized in this LiGeoreference tab are also be found within LiDAR360’s Reproject tool.
When the target horizontal coordinate system’s datum is different from the default WGS84 datum, users must supply the seven transformation parameter values as inputs to the processing routine. If these parameter settings are not already known, the Seven Parameter Solution tool found in LiDAR360 can use the default point cloud output by LiGeoreference in conjunction with as few as three distinct vertical survey control points to calculate the values needed. Once the seven-parameter transformation is applied to the individual point coordinates they can then be reprojected to an alternative horizontal coordinate system. If no datum transformation is required (e.g. when transforming from WGS 84 UTM 13N to WGS 84 UTM 12N) using seven parameter transformation is not necessary.
There are multiple methods that can be used when the target point cloud elevation reference model is something other than the default WGS84 ellipsoid. If the surveying area is small enough it can considered spatially homogenous, meaning the divergence between ellipsoid height and target geoid height is sufficiently small, users can opt to set the Geoid model to ‘Custom’ and simply enter the difference between ellipsoid height and the target geoid model in the dz field to adjust the elevation values of each 3D point by a consistent amount. When the surveyed area is too large to be considered spatially homogenous, users can either choose from the list of alternative elevation datums included in the software (i.e. geoids EGM2008, EGM96, and EGM84) or apply a elevation adjustment based on surveyed vertical control points.
Strip Alignment
The boresight error between laser scanner coordinate system and GNSS/INS coordinate system is the largest systematic error source found in LiAir V data after the post-processing pose estimation (PP POS) routine is run on its trajectory data using LiNav. Strip Alignment tools found in LiDAR360 allow users to automatically calculate the optimal boresight correction values needed to minimize systematic error in the point cloud data sourced from inaccuracies in the calibration parameter values hardcoded into the dev.cal file delivered with each LiAir V unit. When relying on the automated calibration parameter value calculator to correct systematic error in LiAir V trajectory data, flight patterns must be designed in such a way that they include separate flight lines, or strips, that run roughly parallel to or intersect with each other for the strip alignment tool to be used. Inter-strip distances should be small enough to ensure that at least 30% overlap between the areal footprints generated from POS segments that are immediately adjacent to one another.
Using functions found in the Strip Alignment toolset, the LiNav output POS file can be split into separate segments with each one representing a discrete chunk of the time block during which LiDAR Data was collected. Individual point clouds are then generated by georeferencing only LiVox measurements taken between for each strip’s start and end time. A facet matching algorithm is then used to determine the offsets between overlapping point clouds and from these offset values △X, △Y, △Z, △Roll, △Pitch, △Heading boresight correction values will be calculated. By default, only the angle error parameters (△Roll, △Pitch, △Heading) are calculated automatically because they have the largest impact on the total boresight error. If necessary, users can choose to automatically calculate translation parameters (△X, △Y, △Z) as well but this is generally not necessary.
When clipping the LiNav POS file output, or trajectory file, into segments it is advisable to throw out regions of time during which the UAV turns during data collection. LiDAR measurements in these regions of the point cloud tend to have lower accuracy relative to that of the measurements taken while the drone was flying at a fixed heading. Trajectory splitting can be done interactively or automatically using tools found in LiDAR360. POS segments that include portions of the flight route where the UAV operator took off, landed, performed figure-8 calibration maneuvers, or flew to and from the start and of the automated flight path should also be thrown out of the final set of trajectories from which point clouds are generated in subsequent data processing steps.
After separate point clouds have been created for each POS segment, boresight correction delta values can be calculated automatically. Users can review the Alignment Quality Report generated in the Strip Alignment Toolset prior choosing to apply the deltas. Improved boresight calibration parameter values will decrease the Root Mean Square Error between corresponding facets found in separate but spatially overlapping point cloud segments. Maximum vertical residual (error) values should also fall as the spread of residual distance values between corresponding facets tightens.
Once the automatically calculated calibration parameter delta values have been applied to the individual POS segments, between-strip alignment quality should, in many cases, also be visually improved. Before accepting the automatically calculated boresight correction values, users should launch the Profile Editor and inspect regions were adjacent point clouds overlap and confirm that the parameter value adjustments do in fact improve the fit between separate segments.
It helps to render each POS segment’s associated point cloud in a unique color when carrying out visual point cloud alignments. Vertical profiles with small buffer widths (< 5 cm) should be sampled throughout the point cloud overlap area in planar features such as parking lots, flat and pitched roofs, or sidewalks are found.
The Profile Editor’s 3D viewing window allows users to readily preview and measure the effects of applying automatically calculated boresight correction values and confirm that stratification of points belonging to separate strips is decreased along with the vertical spread (distance) of points within a local area. Vertical point spreads displayed in the 3D view of the Profile Editor should be small in areas were smooth planar features like tennis courts, freshly paved roads, etc. can be found in the point cloud. As a general rule of thumb the vertical spread of a multi-strip LiDAR point cloud collected by a properly calibrated LiAir system, with the operator adhering to best practices for data collection and processing described in this article, should be less than or equal to the system’s advertised accuracy times two. So LiAir V users should expect the vertical spread between points located at approximately (within a 0 to 5 cm radius) to be less than or equal to about 10 centimeters.
In some situations, it may be necessary to manually adjust the boresight angles. It has been shown that compared to angle errors, boresight translation errors are of less importance and in most instances may be omitted from consideration during any the manual correction process. The stepwise geometric misalignment method (Zhang et al., 2010) can be followed when modifying the default LiAir V calibration values. This method is based on the relationship between the point clouds of regular objects, e.g. such as building roof planes, and ground truth observations of these features in the LiDAR dataset used for calibration. The Profile tool found in LiDAR360 can be extremely helpful when taking the measurements needed to find optimal boresight correction angles using the stepwise geometric misalignment method.
The report generated with LiDAR360 after the automatic boresite correction calculator is run cannot be used to statistically describe the overall change in alignment quality for overlapping point cloud regions after a manual adjustment is made (applied) to one or more boresight calibration parameter value. In these instances, the Elevation Difference Quality assessment feature found in the Strip Alignment toolset should be used to analyze and report the between strip elevation differences for overlapping regions both before and after any boresite calibration correction is applied. Users can then compare the reports to see if the changed calibration parameter values yield results with lower standard deviation (variance), root mean square, and average residual vertical offset values (dz) between overlapping areas.
Exporting
LiAir V operations terminate with the production of 3D LiDAR point clouds that reference a user defined-coordinate reference system. Spatial data models (e.g. DEM, DSM, contours, etc.) are created from these outputs using functions found in point cloud post-processing software programs like LiDAR360. Specific workflows may require tools not found in LiDAR360’s Framework, Terrain, or Forestry modules and users will need to export their LiAir V data into common file formats (e.g. las/laz, ply, e57, txt, etc.). This can be easily done with functions built into both LiGeoreference and LiDAR360.
Before exporting individual point cloud segments LiDAR360 overlap points can be labeled and removed from each point cloud. This is typically carried out final stage of the Strip alignment workflow when users have the option to classify or delete eliminate point cloud redundancy by classifying and removing points from overlapping point cloud regions. Once overlap points have been identified and kept in/removed from each strip, the individual point cloud segments can be easily merged into a single point cloud before they are exported from LiDAR360.
It is during the point-cloud post processing stages that most of final LiAir V mapping or surveying project deliverables are generated and final data quality is assessed. Refer to this GVI Knowledge Base for more information on how software programs, like LiDAR360 and LiPowerline, can be used to easily generate industry-specific data products from LiAir V point clouds or asses the accuracy of a specific dataset.
Contact us
For additional information on LiAir V or any topics discussed in this Product Article, please contact GVI at info@greenvalleyintl.com.