Introduction
What is a LiDAR Payload?
LiDAR stands for Light Detection and Ranging and has become an increasingly prevalent technology over the past decade. NASA and meteorologists have used LiDAR, now embedded within cell phones with the release of the iPhone 12 Pro. LiDAR is commonly used in surveying, mapping, and inspection applications where it can provide very detailed and accurate point clouds of an environment. These point clouds are a conglomeration of all the scanned points (of which there are typically hundreds of thousands every second) to form a 3-dimensional image of the environment in a matter like that of a pointillist painting. Lidar scanners alone are very good at using light pulses to determine the distance of an object with impressive accuracy. Still, by itself, this is all it can do: determine relative ranges of objects with no concept of where they are in space.
Figure 1. Example Pointillism Painting (https://artincontext.org/pointillism/)
A complete LiDAR payload commonly integrates a GNSS-aided Inertial Navigation System (INS) and a lidar scanner. This provides the data necessary for georeferencing, which transforms the relative data of the LiDAR scanner such that each point has its own location on a geographic coordinate system. Since the GNSS-aided INS is constantly recording the orientation, position, velocity, and timing (OPVT) of itself in a global reference frame, this data can be fused with the LiDAR ranging data to create a directly georeferenced point cloud telling the user the absolute location of every single point in the point cloud. This is known as geospatial data.
Applications
Mapping
LiDAR is a handy tool for mapping projects that require high precision. Mapping is a staple of various industries, including archeology, agriculture, construction, and conservation. Drone-based LiDAR presents the safest, most accurate, and quickest way to generate environmental information. Drone-based LiDAR allows the end user to produce accurate three-dimensional models of large areas rapidly. A great example of such a use is monitoring the progress of a construction site. Drone LiDAR eliminates the need for laborious terrestrial survey teams to cover a large construction site on foot. Instead, the geospatial data gathered by LiDAR allows the construction team to track the progress of buildings under construction efficiently and compare them to the original schematic. [1]
Inspection
In addition to mapping, inspection is another essential application of LiDAR, as it typically requires exact data to be helpful for the end user. A great example of an inspection project that requires precision is a transmission line inspection. As the global power grid expands, the number of long-distance transmission lines increases. These lines can often traverse hazardous or remote terrain and require continuous monitoring to perform maintenance and asset inventory. These conditions pose challenges for terrestrial-based observation methods, which cannot keep up with the operational requirements of these vast grids. Drone LiDAR allows users to cover long transmission lines in a fraction of the time it would take a ground-based survey team to complete. Additionally, LiDAR typically has multiple returns, meaning that users can map the power lines, the terrain beneath it, and the vegetation in the area as a single scan to improve the efficiency of the inspection. This is very important because monitoring vegetation near power lines is essential to avoid potential hazards like foliage overhangs that could fall on lines and cause a disruption. [2]
Surveying
Surveying is the application that emphasizes the importance of accurate data collection of all the applications discussed so far. Surveying requires high accuracy to determine property lines, building locations, road topography, and more. Generally, survey-grade accuracy refers to an absolute accuracy of 10 centimeters. The classical approach to land surveying requires a ground team and a theodolite to determine vertical and horizontal angles between points. While a tried-and-true method, it does have its drawbacks. These surveys take a long time to cover any sizable portion of land, and any potentially hazardous terrain presents significant risks to the ground survey team. [3]
Drone-based LiDAR systems have disrupted the land surveying space with the enhanced capabilities of this advanced technology. A drone makes it much easier to measure topography since it can access large land areas, thus eliminating the need for guesswork or inferencing. Additionally, data is captured substantially faster with drones than by a traditional land survey team and requires fewer human resources to complete. Lastly, drone LiDAR systems reduce long-term costs as data acquisition happens much more efficiently and rapidly than a ground team. [4]
Actionable Data
Point Cloud Classification
A point cloud can create many other deliverables required for a mapping or inspection project. Point cloud classification assigns a series of predefined labels to groups of points in a point cloud. This determines what points belong to what objects in a point cloud. One of the most prevalent examples of point cloud classification is the classification of ground points. Since many LiDAR scanners have several returns, they can penetrate obstacles like vegetation and reach the ground. This classification of ground points allows users to generate valuable deliverables for their customers.
DEM
One example of such a deliverable is a digital elevation model. A digital elevation model uses only the ground classified points to provide a smooth bare earth elevation model. These are often used in hydrologic modeling, terrain stability, and soil mapping. In hydrology, DEMs are essential for coastal hydrologic modeling applications as they help delineate watersheds and calculate flow accumulation and direction. DEMs are also great ways to model terrain stability and allow the user to predict landslides, as shown in the figure below. [5]
Figure 2. DEM Example (Slope Stability & LiDAR — Terrainworks)
Hillshade Models
Another example of essential deliverables is the hillshade model. This standard method of terrain representation allows for the detection of topographic features that may not be readily apparent otherwise. These models are not meant to be aesthetically pleasing but to have maximum contrast to uncover terrain features. They are typically derived from DEMs or digital surface models (DSMs) and look like a hypothetical light source illuminated by the elevation surface. Hillshades are often used as the background for a cartographic map or to detect and predict terrain hazards like landslides. An example of a hillside is shown below. [6]
Figure 3. Example Hillshade Model (https://www.esri.com/esri-news/arcuser/fall-2014/multi-directional-hillshade-makes-your-maps-pop)
WISPR Systems– The Complete Solution Provider
Founded in 2016, WISPR Systems is a Mississippi-based solution and system manufacturer providing rugged, versatile, and reliable commercial drones and various payload options. WISPR Systems has partnered with Inertial Labs to provide end-users with complete surveying solutions that cover the needs of drones, LiDAR, photogrammetry, and their deliverables within the GIS space across various fields and applications. WISPR is not just a provider of these systems. Instead, WISPR provides surveyors with a completely integrated solution and provides the training necessary so that the user can extract valuable and actionable data. WISPR’s extensive expertise in the remote sensing field allows them to guide users seamlessly through a complete workflow. WISPR is a one-stop provider of complete remote sensing solutions, from flight planning to data capture to data processing and deliverables.
Inertial Labs RESEPI XT-32 vs. M2X
Testing Process
Inertial Labs and WISPR partnered to create a wholly integrated drone LiDAR solution using the Ranger Pro with the RESEPI XT-32 and RESEPI M2X. The drone and the navigation system of each complete solution remained the same; the only change was the LiDAR scanner. Below is a comparison of both scanners in a few key parameters.
Device Name | XT-32 | M2X |
Maximum Range (m) | 120 | 300 |
Range Accuracy (cm) | 1 | 1 |
Number of Returns | 2 | 3 |
Data Points Generated Per Return (pts/s) | 640,000 | 640,000 |
Field of View | 360° (H) | |
Beam Divergence | 0.04° (H) | 0.21° (H) |
Weight (Laser Only) | 0.8 kg | 0.49 kg |
Table 1. XT-32 vs M2X Specification Comparison
WISPR sought to test the performance of both systems by flying the same route at three different AGLs: 50 meters, 100 meters, and 150 meters. For reference, the image below shows the comparison between the flight paths of the M2X and XT-32 at 50 meters AGL in Inertial Labs’ PCMasterPro Software Suite.
Figure 4. 50m AGL Flight Path XT-32 (left) vs. M2X (right)
It is also important to note that each flight path was designed to have 50% overlap to ensure complete, high-density coverage. This overlap is indicated by the color-coordinated passes of the path and the corresponding points. One last parameter to note in this instance is that the scanner was cut down to utilize a horizontal field of view of 120 degrees. With these parameters in place, it is essential to see how this is manifested at different AGLs. The figures below show the flight path at 100 meters AGL and 150 meters AGL.
Figure 5. 100m AGL Flight Path XT-32 (left) vs. M2X (right)
Figure 6. 150m AGL Flight Path XT-32 (left) vs. M2X (right)
With fewer passes at a fixed 120-degree field-of-view, it is easier to see the overlap of each successive pass in the flight as the height AGL increases. With features like heavy vegetation, power lines, and buildings, this environment is a phenomenal place to examine many aspects of the performance of each payload.
Data Analysis and Comparison
The point clouds for all six flights are shown in the figure below, and it becomes clear that the AGL plays a significant role in the detail and density of a point cloud. These point clouds are colorized based on altitude and include all points from the respective flights.
Figure 7. M2X (top row) vs. XT-32 (bottom row) at 50, 100, and 150m AGL (from left to right) Full Point Cloud
From a macro perspective, we can see a few apparent differences between point clouds at different AGLs. As the AGL increases, structures requiring precision mapping, like power lines, lose details or are entirely omitted from the point cloud. A critical difference between the M2X and XT-32 comes from the 100m AGL cloud, where the roofs of the buildings begin to get cut off on the XT-32 but remain intact with the M2X.
Another critical parameter of a LiDAR scanner is its ability to penetrate vegetation, allowing users to get data points of the ground beneath the vegetation. As a result, surveyors can get detailed DEMs, hillshade models, and more, even if objects block the scanner’s line of sight. These models can highlight important terrestrial features not visually apparent to the naked eye. A great example of the power of LiDAR is highlighted in the figures below, which compare data at 100m AGL between the XT-32 and the M2X.
Figure 8. M2X (Left) vs. XT-32 (Right) 100m AGL – Point Cloud
Most of Figure 8 contains very heavy vegetation to the point that the ground beneath the trees in this point cloud is not visible. To see the ground for this entire view, let’s classify and display only the ground points from this cloud, effectively removing all points above the ground surface. Figure 9 below displays the ground-classified cloud.
Figure 9. M2X (Left) vs. XT-32 (Right) 100m AGL – Ground Points
Despite heavy vegetation, these scanners performed exceptionally well to penetrate the ground, even at 100m AGL. Upon inspection, both scanners are highly comparable in this facet, with the M2X having an edge in point density. They also allow us to get a surface model of the cloud in this view, as shown in Figure 10 below.
Figure 10. M2X (Left) vs. XT-32 (Right) 100m AGL – Surface Model
This surface is especially impressive because it uncovered this small ravine running through the center of the landscape that was not visually apparent without using this model. These impressive ground classified clouds were not specific to this view and AGL; the figures below display this capability of both scanners over varying scenarios.
Figure 11. M2X (Left) vs. XT-32 (Right) 150m AGL – Point Cloud
Figure 12. M2X (Left) vs. XT-32 (Right) 150m AGL – Ground Points
Figure 13. M2X (Left) vs. XT-32 (Right) 150m AGL – Surface Model
The example above covers a similar area but at an AGL of 150 meters. This height starts to strain the range of the XT-32, and it shows when we look at the point density of the ground points and surface model. The M2X produces a much more detailed solution in this instance, which makes sense as the range specification of the M2X is 300m and the XT-32 specification is 120m.
We can further illuminate features by deriving a hillshade from the DSMs found in the previous figures. The flight scanner and the AGL can play a role in the detail and accuracy of the hillshade model. The figure below compares the same hillsides at varying AGLs for each scanner.
Figure 14. M2X (top row) vs XT-32 (bottom row) at 50, 100, and 150m AGL (from left to right) Hillshades
As shown above, the M2X produces detailed, clean hillshade models up to 150-meter AGLs. The XT-32 also produces clean hillshade models within its specified range, but once the AGL reaches above 150 meters, we see that part of the data is cut off. Additionally, both scanners allow the user to identify a ravine in this terrain despite it not being visible to the naked eye or on the raw point cloud data. This is partly due to the nature of hillsides, as they are high-contrast models.
In the remote sensing industry, many different terms get thrown around about accuracy. In general, any accuracy specification can fall under either relative accuracy or absolute accuracy. In the context of a complete LiDAR payload, relative accuracy measures the accuracy between points relative to each other within a single project. Absolute accuracy measures how close a measured value is to a known, surveyed location (actual value) in a geographic coordinate system. An example of how relative accuracy and absolute accuracy are measured is shown in the figure below. This is an important distinction when analyzing the quality of LiDAR remote sensing solutions, as these values speak to the quality of different components of the remote sensing payload. Relative accuracy is more dependent on the LiDAR scanner, while absolute accuracy is more dependent on the quality of the inertial navigation systems (INS) on board. Both parameters are essential for surveying applications to get the job done correctly.
Figure 15. Relative vs Absolute Accuracy
The first accuracy parameter that was examined was relative vertical accuracy. This was done by taking data from 2 parallel passes and comparing a slice of the overlapping section. Datapoints are colored based on selection, so what points came from what selection will be evident. An example of this parallel pass method at 50 meters AGL is shown in the figure below.
Figure 16. Parallel Passes at 50 meters AGL
All data used for relative accuracy measurements were collected on flat ground. Data was collected and compared at 50 meters AGL, 100 meters AGL, and 150 meters AGL.
Figure 17. M2X (Top) vs XT-32 (bottom) 50m AGL Relative Accuracy Comparison
Figure 18. M2X (Top) vs XT-32 (bottom) 100m AGL Relative Accuracy Comparison
Figure 19. M2X (Top) vs XT-32 (bottom) 150m AGL Relative Accuracy Comparison
As shown in the above figures, it is evident that the M2X and XT-32 perform almost identically in relative accuracy for 100 meters AGL or less comparisons. Both devices have close to 3 cm relative accuracy at 50 meters AGL and close to 6 cm relative accuracy at 100 meters AGL. We see that at 150 meters, the M2X has a more significant point density than the XT-32 and has a relative accuracy of about 9 cm, while the relative accuracy of the XT-32 is closer to 10 cm.
The next step is to check the vertical absolute accuracy of the system. This was done using ground control points (GCPs), surveyed points of a known location. This environment contained 7 GCPs, and in the Lidar360 point cloud analysis software, the LiDAR points from both scanners were compared to the surveyed locations. The de-bias does not delete, add, or change the relative position of the points; it is only a systematic adjustment/shift of the entire point cloud based on observed/detected systematic offset. In this instance, a manual shift of 0.151 feet (~4.6cm) was applied to the cloud. Table 2 below compares the RESEPI M2X and RESEPI XT-32 in crucial metrics of absolute accuracy at an altitude of 60 meters.
Parameter | RESEPI M2X | RESEPI XT-32 |
Average Magnitude | 0.045ft (1.37cm) | 0.042ft (1.28cm) |
Standard Deviation | 0.055ft (1.68cm) | 0.055ft (1.68cm) |
Root Mean Square | 0.051ft (1.55cm) | 0.051ft(1.55cm) |
Average dz | 0.000ft | 0.004ft(0.12cm) |
Minimum dz | -0.065ft (-1.98cm) | -0.058ft(1.77cm) |
Maximum dz | 0.088ft (2.68cm) | 0.092ft(2.8cm) |
Table 2. Absolute Accuracy Comparison
The table above shows that both payloads had impressive absolute accuracies after debiasing with a root mean square (RMS) accuracy of less than 2 cm. It is readily apparent that all absolute accuracy metrics are well within the survey-grade benchmark of 10 cm. Most of them are within 1/5th of the required threshold. The similarity of both payloads’ absolute accuracy points to the repeatability of Inertial Labs’ inertial navigation systems, as both absolute accuracies are nearly identical. Regardless of the scanner, both devices can give surveyors immense confidence in their results after they survey an environment.
Conclusion
LiDAR is a powerful technology that allows users to capture an environment in detail and with high accuracy while saving man-hours and reducing the safety risk of the survey team. LiDAR payloads can produce insightful deliverables such as DEMs, DSMs, and hillshade models that give the end user actionable results. WISPR systems provide a wholly integrated drone-LiDAR system and have the expertise to support users in getting the most out of their data.
The RESEPI-M2X is the higher-end model of the scanner comparison, and it is understandable when looking at head-on comparisons at higher AGL heights. It provides dense models and impressive vegetation penetration, allowing users to fly at various AGLs while maintaining a high-quality data precision and density standard. The XT-32 proved to be a fierce competitor to the M2X, especially at AGL heights of 100 meters or less. The XT-32 is a cost-effective model that can still provide dense, accurate point clouds of many environments, penetrate vegetation, and provide accurate, actionable deliverables to customers.
Ultimately, the suitable scanner for the end user depends on several factors: budget, environment, accuracy requirements, etc. Both scanners could produce accurate dense clouds, penetrate heavy vegetation, and generate insightful DSMs, DEMs, hillshades, and more. For applications at lower AGL, such as asset inspection, the XT-32 may be the right choice to get detailed imagery of essential resources. If a user needs to map out a large stretch of land, then the M2X could allow the user to gather data from a higher AGL while maintaining adequate accuracy and density for many mapping projects. No matter what, users can’t go wrong when trusting WISPR systems to provide a complete drone-LiDAR system that reliably produces accurate, actionable data in a user-friendly fashion.
Bibliography
[1]“A Guide To Using Drones and LiDAR Technology for GIS Mapping,” www.duncan-parnell.com. https://www.duncan-parnell.com/blog/using-drones-for-gis-mapping/
[2]Benowitz, “LiDAR Equipped UAVs,” enterprise-insights.dji.com, Jul. 28, 2022. https://enterprise-insights.dji.com/blog/lidar-equipped-uavs
[3] 2022 “Land Surveying: The Process and the Tools,” EngineerSupply. https://www.engineersupply.com/land-surveying.aspx
[4]“How Have LiDAR Drones Transformed Traditional Land Surveying?,” blog.smartdrone.us, Sep. 05, 2022. https://blog.smartdrone.us/insights/how-have-lidar-drones-transformed-traditional-land-surveying#:~:text=LiDAR%20drones%20enable%20drone%20surveyors%20to%20map%20the (accessed Dec. 08, 2023).
[5] GISGeography, “DEM, DSM & DTM Differences – A Look at Elevation Models in GIS,” GIS Geography, Mar. 09, 2016. https://gisgeography.com/dem-dsm-dtm-differences/
[6} 2018 Čučković, “Some thoughts on hillshade models for Lidar analysis,” landscapearchaeology.org, Nov. 10, 2018. https://landscapearchaeology.org/2018/lidar-hillshade/ (accessed Dec. 08, 2023).
For more information:
Anton Barabashov
VP of Business Development
Inertial Labs Inc.
sales@inertiallabs.com