Posts

Understanding the Role and Impact of Each on your UAV-LiDAR Missions

This blog (second in a two-part series) continues our explanation of the variables which influence boresight calibration and strip alignment when flying drone-based LiDAR missions. If you missed the first part of this series, we recommend you check it out before continuing with this blog.

LiDAR boresight calibration aims to calibrate the LiDAR boresight angle, as demonstrated in Figure 1 below. Boresight calibration addresses the issue of alignment of the LiDAR and the IMU body frames – the correlation between the LiDAR boresight angles and the INS attitude. Some specialized third-party software can decorrelate these two error sources and estimate/calibrate the LiDAR boresight angles. However, some software developers list these capabilities but simply lump all errors together and employ automated algorithms to align the point clouds.

Figure 1. LiDAR boresight calibration – before (left) and after (right

Figure 1. LiDAR boresight calibration – before (left) and after (right

Strip alignment (also known as strip adjustment) can help improve the consistency between several adjacent UAV-LiDAR flight lines. Strip alignment workflows utilize algorithms which align the LiDAR point clouds from one strip to another without necessarily knowing the source of the error. Certain LiDAR post-processing software options will have built-in capabilities for strip alignment. These ensure alignment of the point cloud based on common features captured in overlapping data sections and computing the geometric transformation between these datasets. Addressing boresight error is typically the first step in the strip alignment workflow.

Strip alignment procedures can be used if the system has undergone boresight calibration but there are remaining residual discrepancies in overlapping areas that cannot be explained by other sources of error. In other words, strip alignment is only necessary if there are still minor misalignments noticeable after the LiDAR point cloud has been georeferenced. Factors such as flight altitude profile, manual flight, wind or other environmental conditions can impact the flight trajectory in subtle ways that keep the point clouds from aligning as intended.

As another example, let’s consider flying a Geo-MMS LiDAR payload over a construction site. We fly the mission, collect and process the data and visualize it in our 3D point cloud viewer. Upon visual inspection, we see two lines in the point cloud for a feature we know should only be represented by one line (e.g., edge of a building roof). This may have been the result of poor flight planning and execution, inexperienced piloting or flying in harsh weather conditions. If you are on a tight schedule, strip alignment software can make a positive difference by offering different options to improve the consistency between flight lines. Different algorithms can correct the position and angles of the drone across the mission duration to align the data as it was intended to be collected originally. Post-processing the data in this fashion can save time by avoiding a repeat mission.

Figure 2. Misalignment of adjacent LiDAR strips intersecting a feature of interest (left) and with automated strip alignment algorithms employed (right).

Figure 2. Misalignment of adjacent LiDAR strips intersecting a feature of interest (left) and with automated strip alignment algorithms employed (right).

In the example described above, the alternative to owning strip alignment software would be a repeat mission (correct planning and execution in better flying conditions). Repeating a mission as a one-off is fine, but users who regularly have to re-fly missions may stand to benefit from a dedicated point cloud strip adjustment software.

Software such as BayesStripAlign from BayesMap Solutions focus on automated strip alignment for LiDAR point cloud data from UAVs and manned aircraft. If correct flight planning and execution are adhered to, it is rarely necessary to add this level of capability to UAV-LiDAR post-processing, given the onboard INS is of the highest quality. Specialized software can be useful when performing work at high altitudes from manned aircraft with long-range sensors such as the Teledyne Optech CL-360XR. For LiDAR pilots who are forced to operate in harsh conditions (something we advise against), a strip alignment software can help correct for the environmental impacts exerted on your data collection methods. However, most software have limited support for UAV-LiDAR operations. The most practical approach is to avoid operating the system in harsh weather, in which all sources of the aforementioned error sources will exceed normal ranges. To counter this, Geodetics developed a set of algorithms that can adaptively adjust the boresight angles for each individual LiDAR strip, eventually aligning all strips together. The figure below demonstrates system operation from aligning two sets of overlapping point clouds based on adjustment of the boresight angles. The algorithm converges rapidly and estimates accurate misalignment angles. Future developments will extend this algorithm to more generic cases of data collection scenarios in which the AOI is primarily featureless terrain.

Figure 3. Example taken from Geodetics’ boresight angle adaption software.

Figure 3. Example is taken from Geodetics’ boresight angle adaption software.

Geo-MMS customers can rest assured knowing that the proprietary Defense-grade inertial navigation technology developed by our navigation scientists and engineers is amongst the highest quality in the industry.

Originally published by Geodetics 

Understanding Two Critical Principles of LiDAR Data Acquisition

If you are familiar with any type of LiDAR data acquisition, you have likely stumbled across terms including ‘boresight calibration’ and ‘strip alignment’ (or strip adjustment). These terms relate to separate strategies for adjusting and aligning adjacent strips of aerial LiDAR data. In this blog (first in a two-part series), we will address why boresight calibration and strip alignment are important considerations for UAV-LiDAR missions. In part two of this series, we will provide practical information to the LiDAR end-user which further addresses and explains the points raised in this blog.

Boresight calibration and strip alignment are particularly prevalent for UAV-based LiDAR solutions such as Geo-MMS LiDAR. UAV-LiDAR is acquired through many parallel ‘strips’ of LiDAR which overlap adjacent strips slightly (< 10%). The purpose of boresight calibration is to correct for minute misalignments between adjacent strips noticeable in the raw point cloud.

Figure 1. Flat terrain, captured across seven parallel UAV-LiDAR flight lines.

The main concerns for the LiDAR end-user are:

  • Do we need data processing procedures to adjust our point clouds for possible misalignments?
  • Is additional software needed to make these adjustments?

But first, we should address what causes these misalignments in the first place. Once the error sources are identified, we can better understand the principles behind LiDAR boresight calibration and strip alignment. When looking at the error budget of any aerial LiDAR system, we experience both systematic and random errors. With correct system calibration, systematic errors are removed and random errors are minimized.

Systematic errors result in the systematic deviation of laser footprint coordinates. Integration of Geo-MMS LiDARor Point&Pixel payloads on your drone requires the axis of the scanning reference coordinate system and inertial platform reference coordinate system to be parallel. While physically mounting the payload to the UAV, it is not guaranteed that they will be parallel (human hands are not precision instruments). This results in what we term boresight error.

Figure 2 below summarizes the geometry of different components in the Geo-MMS system in a simplified vector-based visual. Once the LiDAR sensor fires a pulse measuring the range from the laser to the scanned feature, , we need three more observations to georeference the captured point:

  • The position, provided by the GPS () and represented in the PPK solution
  • The leverarm offset to transform the GPS position to the IMU center, marked as ()
  • The boresight shift to transfer the position from the IMU center to the LiDAR sensor center, marked as ()

Figure 2. Direct georeferencing geometry

From vector geometry, the global position of the point, can be determined by adding the sequences of four vectors:

In this geometry reconstruction, we consider the translation between the components. Besides translations, the rotation between these components is critical. Considering the rotational influence at different sensor frames, Equation (1) can be rewritten as:

  • represents the rotation matrix, reconstructed based on the roll, pitch, and yaw of Geodetics’ Navigator (INS)
  • represents the rotation matrix from the IMU to the LiDAR, represented as the LiDAR boresight angle

Now that we have covered the basic geometry, we can highlight the main contributory points that direct georeferencing has on practical LiDAR usage:

  • refers to the GPS-to-IMU leverarm This is a known calibrated value and is pre-configured with your Geo-MMS system.
  • refers to the IMU-to-LiDAR offset or LiDAR boresight shift. This too is a known calibrated value and is pre-configured with your Geo-MMS system.
  • refers to the LiDAR boresight angle. This too is pre-configured with your Geo-MMS system.
  • refers to the PPK solution of the GPS antenna.
  • refers to the INS-based rotation. This is determined by the INS (roll, pitch, yaw) and its accuracy is typically a function of the IMU grade, dual-GPS based heading and filter-tuning procedures.

The boresight error between the LiDAR sensor and the onboard GPS/INS coordinate system is the largest systematic error source in UAV-LiDAR. The laser footprint error produced as a result of boresight error is also impacted by flight altitude (AGL) and scan angle.

The main purpose of this geometry breakdown is to highlight the five most important variables which directly impact the quality of the LiDAR data generated. Among these variables, two cannot be pre-calibrated: PPK solution and INS attitude. Several algorithms are implemented in Geodetics’ LiDARTool software to improve the PPK and INS solutions including forward-backward PPK smoothing and rounding point coordinate values to fine resolution grids. However, leftover errors can still impact point cloud reconstruction quality in harsh environments.

In part two of this series, we will more closely examine the questions posed at the beginning of this blog, pertaining to the two main concerns of the LiDAR end-user.

Originally posted by Geodetics