Available via license: CC BY 4.0
Content may be subject to copyright.
EVALUATION OF A GNSS/IMU/LIDAR-INTEGRATION FOR AIRBORNE LASER
SCANNING USING RTKLIB PPK AND PPP GNSS SOLUTIONS
F. Pöppl1∗
, G. Mandlburger1, N. Pfeifer1
1Technische Universität Wien, Vienna, Austria - (florian.poeppl, gottfried.mandlburger, norbert.pfeifer)@geo.tuwien.ac.at
KEY WORDS: Direct georeferencing, sensor orientation, sensor integration, trajectory estimation
ABSTRACT:
Airborne laser scanning allows for efficient acquisition of accurate 3D data for large areas. Because georeferencing of the LiDAR
data requires knowledge of the platform trajectory, the laser scanner system commonly comprises a global navigation satellite
system (GNSS) receiver/antenna and an inertial measurement unit (IMU). The standard processing pipeline consists of GNSS/IMU
integration, georeferencing, and subsequent adjustment of the laser data. Here, we consider a holistic GNSS/IMU/LiDAR-integration
approach based on least-squares adjustment. The GNSS is loosely coupled, and the GNSS positions are obtained using either post-
processing kinematic or precise point positioning GNSS processing strategies using the open-source software RTKLib. In this
contribution, we compare the resulting point clouds to those of a standard processing workflow and evaluate the impact of the
different processing strategies on point cloud quality in terms of internal consistency and absolute accuracy for a airborne laser
bathymetry (ALB) dataset. Although the GNSS solutions themselves differ strongly, both the PPK- and the PPP-derived point clouds
show better strip differences (below 2.5cm) and similar absolute accuracy (<4 cm RMSE w.r.t. reference targets after correction of
constant datum shift) compared to the reference solution.
1. INTRODUCTION
Laser scanning is a widely used surveying technique based on
light detection and ranging (LiDAR) in conjunction with a scan-
ning mechanism. With the laser scanner mounted on a moving
platform, large areas may be efficiently mapped and digitally
represented as 3D point clouds. Here, we specifically consider
airborne platforms. The laser scanner measurements consist of
a range measurement together with one or two angular measure-
ments. To transform the ranging and angle measurements of the
laser scanner into a georeferenced coordinate system, the posi-
tion and orientation of the scanner are required. Commonly, a
global navigation satellite system (GNSS) receiver and antenna
as well as an inertial measurement unit (IMU) are integrated
with the laser scanner, which allows computing the required
trajectory (Pöppl et al., 2023a).
The industry-standard approach is fusion of GNSS and IMU data
using a Kalman filter, possibly followed by strip adjustment to re-
duce discrepancies between overlapping flight strips (e.g., Glira
et al. 2015 and Jonassen et al. 2023). Various strategies ex-
ist for GNSS and IMU integration (Groves, 2013), most com-
monly: loose coupling (using pre-processed GNSS positions),
tight coupling (using raw GNSS code and carrier phase measure-
ments), or deep coupling (e.g., inertial aiding of GNSS tracking
loop). While the tight coupling of GNSS and IMU is generally
expected to be more accurate and robust compared to a stand-
alone GNSS solution, a loose coupling of GNSS significantly re-
duces the complexity of the integration architecture and allows
interchangeable use of different GNSS processing modes and
software solutions. Strip adjustment, as an additional improve-
ment of the trajectory subsequent to the GNSS/IMU integra-
tion, may be seen as loose coupling of LiDAR to the navigation
sensors.
In this contribution, we consider a sensor fusion architecture
based on loose coupling of GNSS and tight coupling of IMU
∗Corresponding author
and LiDAR. All measurements are considered simultaneously
in a joint least-squares adjustment procedure (see Fig. 1), which
is derived from the more general maximum a-posteriori (MAP)
problem (Pöppl et al., 2023a). Similar methodology has already
been applied in robotics applications (Cadena et al., 2016), pho-
togrammetry (Cucci et al., 2017) and laser scanning (Brun et
al., 2022). The advantage of tightly coupling IMU and LiDAR
is the possibility of correcting short-term trajectory errors with
less risk of deforming the point cloud than with classical strip
adjustment which disregards the underlying navigation data.
In our recent work, we presented a GNSS/IMU/LiDAR-integra-
tion procedure for true direct georeferencing(i.e., without ground
control) of airborne laser scanning data, together with possible
ways of accounting for time-correlated GNSS errors in such a
loose coupling (Pöppl et al., 2023b). We continue along the
same line of investigation and evaluate (1) the use of loosely
coupled GNSS in a joint GNSS/IMU/LiDAR-integration, (2)
the suitability and accuracy of GNSS processing using the open-
source RTKLib software in a high-end surveying application
and (3) the differences between a differential post-processed
kinematic (PPK) and a precise point positioning (PPP) solution.
This evaluation is performed for an airborne laser bathymetry
(ALB) dataset acquired with a high-end laser scanner, and
compares an industry-standard workflow with our GNSS/IMU/-
LiDAR-integration, the latter processed separately with PPK-
and with PPP-GNSS. Note that this is not a discussion of the
technical details of PPK and PPP processing, but rather an at-
tempt to analyze the the practical aspects of the RTKLib PPK-
and PPP-solutions with respect to the GNSS solutions them-
selves, the estimated platform trajectory as well as the internal
consistency and absolute accuracy of the final point cloud.
In Section 2, we describe the processing and evaluation meth-
odology. In Section 3, we present the dataset used and evaluate
the results, followed by a discussion and outlook in Section 4.
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLVIII-1/W3-2023
2nd GEOBENCH Workshop on Evaluation and BENCHmarking of Sensors, Systems and GEOspatial Data
in Photogrammetry and Remote Sensing, 23–24 October 2023, Krakow, Poland
This contribution has been peer-reviewed.
https://doi.org/10.5194/isprs-archives-XLVIII-1-W3-2023-161-2023 | © Author(s) 2023. CC BY 4.0 License.
161
NLS
adjustment
GNSS
positions
IMU raw
measurements
LiDAR point
measurements
Trajectory &
calibration
Georeferencing
Plane extraction
& matching
LiDAR
correspondences
Initial trajectory
Kalman filter
Point cloud
PPK or PPP
Figure 1. Processing workflow for the combined
GNSS/IMU/LiDAR integration.
2. PROCESSING METHODOLOGY
The methodology for integrating the GNSS, IMU and LiDAR
measurements is based on MAP estimation. Measurements
yi, i = 1, . . . , n are modelled as explicit functions of the
(unknown) parameters xplus noise
yi=fi(x) + ϵi,
where the measurement errors ϵiare assumed to follow Gaussian
distributions N(0,Σi). Thus, the optimal (in the MAP sense)
parameters are given by the non-linear least squares (NLS) es-
timator
x
⋆= argmin
x
n
X
i=1
(fi(x)−yi)TΣ−1
i(fi(x)−yi).(2.1)
In other words, the least-squares estimate x⋆minimizes the
difference between model predictions fi(x)and measurements
yifor the GNSS, IMU and LiDAR measurements.
For the specific measurement equations as well as the model-
ling of time-correlated GNSS errors, refer to Pöppl et al. (2023a)
and Pöppl et al. (2023b). The IMU measurements are used as
output by the sensors, with no additional pre-processing. The
LiDAR correspondences are derived from an initial georeferen-
cing (cf. Fig. 1) using voxel-based plane extraction and match-
ing. The GNSS measurement equations are adapted for use
with a gyro stabilization mount by accounting for the GNSS an-
tenna position, which varies with respect to the IMU coordinate
system.
In a loosely coupled GNSS integration, the raw GNSS meas-
urements are pre-processed and the derived antenna positions
are included in the adjustment as observations. Here, we look
at the suitability of using RTKLib for GNSS processing in the
context of airborne laser scanning with high-end sensors. RTK-
Lib (Takasu et al., 2009) is an open-source GNSS processing
software capable of static and kinematic RTK, PPK and PPP
processing. RTKLib is widely used in research, especially ro-
botics (e.g., Li et al. 2020), but also integrated in commercial
GNSS products such as Emlid Studio (Emlid, 2023). RTK-
Lib is used here in two modes, PPK and PPP, to produce two
different solutions. In both cases, all available constellations
were used (GPS, Galileo, GLONASS, Beidou) together with
final precise GNSS products and a 15 deg elevation mask. For
the PPK processing, RTKLib’s Kalman filter was run only in
forward direction due to difficulties with ambiguity resolution
in the combined forward-backward processing. For the PPP
processing, the current RTKLib versions (2.4.2+) are capable
of unstable and experimental ambiguity resolution (PPP-AR,
Takasu 2013). Gurturk et al., 2022 investigated the accuracy
of RTKLib’s (among others) PPK and PPP with ambiguity res-
olution for two aircraft flights, with PPP-AR achieving a posi-
tioning accuracy of 6 cm compared to a PPK reference solution.
However, we were unable to obtain any satisfactory results with
RTKLib’s PPP-AR and therefore use the demo5 version (Ever-
ett, 2022b) where ambiguity resolution is disabled for PPP. The
demo5 version of RTKLib is primarily developed for low-cost
receivers (Everett, 2022a), but has generally been more usable
and reliable than the standard version in our testing. However,
without ambiguity resolution, PPP is expected to be less accur-
ate compared to the results from Gurturk et al. (2022). A generic
antenna profile is available for the antenna, although use of this
profile yielded no noticeable improvement - suggesting that the
calibration may be inappropriate for this specific antenna unit
or due to effects of the aircraft fuselage (cf. Rao 2013, Ch. 5).
The two resulting GNSS solutions are used separately as input in
the adjustment procedure (Fig. 1) to obtain two trajectories. For
each trajectory, an initial georeferencing is performed and the
resulting point clouds are used (separately) for extracting and
matching planar features. Then, the adjustments are repeated,
now with GNSS, IMU and LiDAR measurements. The final
trajectories and georeferenced point clouds are then evaluated.
3. DATA & RESULTS
The evaluation is carried out for a RIEGL VQ-880-GII airborne
laser bathymetry (ALB) system, a high-end topo-bathymetric
laser scanner integrated with an Applanix AP60 GNSS/IMU
on a gyro stabilization mount. This ALB dataset was acquired
in February 2023 in the area of the pre-Alpine Pielach river,
near Loosdorf, Lower Austria (Fig. 2). In this area, high-quality
reference data is available and will serve as ground truth for eval-
uating absolute accuracy. The reference data was acquired by
RTK survey, with an expected accuracy of 1.5 cm in planimetry
and 4 cm in altimetry1.
Data from both the infrared and the green channel is included in
all analyses. For the PPK processing, a virtual reference station
(VRS) in the center of the study area is used due to the lack
of close-by physical base station. In contrast, the advantage
of PPP is of course the complete independence of any local
infrastructure.
Fig. 3 shows the differences between the estimated GNSS an-
tenna positions of the PPK- and PPP-solutions. The two solu-
tions differ by up to 20 cm, especially for the height component.
In addition to the time-varying discrepancies, a constant offset
is present between both solutions (Table 1). After the adjust-
ment process, the final trajectories are again compared (Fig. 4).
The constant offset remains but the time-varying differences are
reduced. Within the scan strips the discrepancies are reduced
to below 2.5cm due to the LiDAR-derived observations. Out-
side of the flight strips, i.e., when the airplane turns and no
scan data is acquired, the differences largely remain (see Fig. 5
and Fig. 6). Note also that the difference becomes less noisy
after adjustment due to the introduction of the (same) IMU
measurements in both adjustments, which smooth out some of
the higher-frequency components.
In Fig. 7, the final point clouds of three processing runs are com-
pared: a) the proprietary workflow of tightly coupled GNSS/IMU
processing using Applanix’s POSPac software, followed by strip
adjustment using RIEGL’s RiPRECISION, b) our GNSS/IMU/-
LiDAR-integration with the PPK-GNSS positions and c) our
1According to the RTK service provider.
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLVIII-1/W3-2023
2nd GEOBENCH Workshop on Evaluation and BENCHmarking of Sensors, Systems and GEOspatial Data
in Photogrammetry and Remote Sensing, 23–24 October 2023, Krakow, Poland
This contribution has been peer-reviewed.
https://doi.org/10.5194/isprs-archives-XLVIII-1-W3-2023-161-2023 | © Author(s) 2023. CC BY 4.0 License.
162
2000 0 2000 4000 6000 8000
X (East) [m]
2000
0
2000
4000
6000
Y (North) [m]
VRS
Trajectory
Figure 2. Platform trajectory and ALS point cloud colored by
reflectance. Coordinates are given in a locally defined Cartesian
coordinate system with east-north-up (ENU) axes.
Shift x / y / z [cm]
GNSS (before adjustment) -5.78 / 1.46 / -1.06
Trajectory (after adjustment) -5.89 / 2.22 / -0.54
Table 1. Average difference between PPK- and PPP-solutions.
GNSS/IMU/LiDAR-integration with the PPP-GNSS positions.
The strip differences are used as a measure of the internal
consistency of the point cloud; here, the proposed adjustment
method (both the PPK and the PPP variant) performs slightly
better than the standard workflow, with strip differences reduced
to below 2.5 cm (excepting vegetation). The distance to the ref-
erence data is determined by first selecting points within 1 m
of the reference target, then estimating a best-fitting plane and
computing the normal distance from the reference target’s po-
sition to the plane. In terms of absolute accuracy, all three
solutions exhibit a (different) datum shift with respect to the
reference targets. Because the reference targets are made up of
horizontal ground control points and sloped rooftops of varying
azimuth, the 3-parameter (X/Y/Z) datum shift may be recovered
by minimizing the root mean squared error (RMSE) of normal
distance between point cloud and ground truth (i.e., by linear
least-squares). Table 2 shows the estimated datum shifts for all
three solutions, as well as the RMSE after application of the
datum shift. The datum shifts are different, but of similar mag-
nitude (∼10 cm) and the RMSE (after shifting) for all solutions
is below 4cm. Note that while the PPK- and PPP-GNSS solu-
tions initially show time-varying differences of up to 20 cm to
50800 51000 51200 51400 51600 51800 52000 52200
Time [s]
10
0
10
GNSS difference [cm]
x y z
Figure 3. Differences in position between the PPK- and the
PPP-processed GNSS solution, over time. The mean differences
(cf. Table 1) are shown as dashed lines.
50800 51000 51200 51400 51600 51800 52000 52200
Time [s]
10
0
10
Trajectory difference [cm]
x y z
Figure 4. Differences in position between the final trajectories
after adjustment using PPK-GNSS and after adjustment using
PPP-GNSS, over time. Time periods marked in gray correspond
to the flight strips (i.e., where LiDAR data is acquired). The
mean differences (cf. Table 1) are shown as dashed lines.
Shift x / y / z [cm] RMSE [cm]
Proprietary -3.32 / 5.26 / -6.82 3.98
PPK-adjustment 0.74 / 8.93 / 1.76 3.76
PPP-adjustment -5.10 / 11.46 / 1.31 3.74
Table 2. Datum shift and RMSE for all three solutions, derived
by fitting the point clouds to the reference targets. Note that the
point cloud shifts for the PPK- and the PPP-adjustment differ by
-5.85 / 2.53 / -0.44 , which matches the average difference
between the trajectories in Table 1.
each other, this reduces to a maximum difference of 2.5 cm after
adjustment for both trajectory and point cloud. Unsurprisingly,
the initial constant difference in the GNSS positions (Table 1)
carries over to the datum shift (Table 2). This implies that while
short-term errors can be corrected by inclusion of the LiDAR
data, long-term effects (including constant shifts) cannot.
The height difference between the PPK and the PPP blocks are
shown in Fig. 8, and are below 2.5cm. The block differences
correlate highly with the position differences of the adjusted
trajectory, with the larger differences in the locations where the
trajectory solutions show a corresponding difference in posi-
tion (see Fig. 6). A closer look at the area marked with the black
rectangle in Fig. 8, where the block difference reaches 2cm and
suitable reference targets are available, confirms a positional
shift affecting the whole area which is also reflected in the ref-
erence target distances (see Fig. 9). Specifically, the locations
of all ground control points, as determined from the respective
point clouds, are shifted in height by 1-1.5 cm. Conversely in
the area of negative block difference, the ground control points
are shifted in the other direction. Therefore, it is unclear which
solution is superior as both strip differences and RMSE w.r.t.
reference are the same for PPK and PPP. But, both show similar
or better point cloud quality than the proprietary solution.
The constant datum shift between the PPK-solution and the ref-
erence data may have different reasons. It is possible that GNSS
processing errors, such as insufficient treatment of atmospheric
effects, may account for some of it. Since different base sta-
tions were used for the acquisition of the reference data (using
RTK-GNSS) and the kinematic processing (PPK), a possible
datum error somewhere along the processing chain cannot be
fully excluded. The RMSE after application of the datum shift
is in line with the expectations for the accuracy of the reference
data, but the constant offset is likely systematic. Still, this is not
limited to the RTKLib processing, as a different constant off-
set of similar magnitude was determined when processing with
the industry-standard proprietary workflows of the navigation
system and laser scanner manufacturers.
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLVIII-1/W3-2023
2nd GEOBENCH Workshop on Evaluation and BENCHmarking of Sensors, Systems and GEOspatial Data
in Photogrammetry and Remote Sensing, 23–24 October 2023, Krakow, Poland
This contribution has been peer-reviewed.
https://doi.org/10.5194/isprs-archives-XLVIII-1-W3-2023-161-2023 | © Author(s) 2023. CC BY 4.0 License.
163
2000 0 2000 4000 6000 8000
2000
0
2000
4000
6000
x
2000 0 2000 4000 6000 8000
2000
0
2000
4000
6000
y
2000 0 2000 4000 6000 8000
2000
0
2000
4000
z
-10 cm
-8 cm
-6 cm
-4 cm
-2 cm
0 cm
2 cm
4 cm
6 cm
8 cm
10 cm
Figure 5. De-meaned differences in position between the PPK- and the PPP-processed GNSS solution for X, Y and Z axes (ENU).
2000 0 2000 4000 6000 8000
2000
0
2000
4000
x
2000 0 2000 4000 6000 8000
2000
0
2000
4000
y
2000 0 2000 4000 6000 8000
2000
0
2000
4000
z
-2.5 cm
-2.0 cm
-1.5 cm
-1.0 cm
-0.5 cm
0.0 cm
0.5 cm
1.0 cm
1.5 cm
2.0 cm
2.5 cm
Figure 6. De-meaned differences in position between the trajectory obtained from the PPK-GNSS/IMU/LiDAR adjustment and the
trajectory obtained from the PPP-GNSS/IMU/LiDAR adjustment, for X, Y and Z axes (ENU). In comparison to Fig. 5, this depicts the
remaining time-varying differences in platform position after exploiting the overlaps in the LiDAR data to correct trajectory errors.
0 1000 2000 3000 4000 5000
0
1000
2000
3000
4000
GCPs Roofs
(a) Proprietary solution.
0 1000 2000 3000 4000 5000
0
1000
2000
3000
4000
GCPs Roofs VRS
(b) PPK-based solution.
0 1000 2000 3000 4000 5000
0
1000
2000
3000
4000
GCPs Roofs
(c) PPP-based solution.
-10 cm
-8 cm
-6 cm
-4 cm
-2 cm
0 cm
2 cm
4 cm
6 cm
8 cm
10 cm
Figure 7. Maximum pairwise height differences of all strips and normal distances to reference targets. The 98 reference targets are
made up of approximately 55 horizontal ground control points and 43 sloped rooftops. Their coordinates are shifted slightly for better
visualization.
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLVIII-1/W3-2023
2nd GEOBENCH Workshop on Evaluation and BENCHmarking of Sensors, Systems and GEOspatial Data
in Photogrammetry and Remote Sensing, 23–24 October 2023, Krakow, Poland
This contribution has been peer-reviewed.
https://doi.org/10.5194/isprs-archives-XLVIII-1-W3-2023-161-2023 | © Author(s) 2023. CC BY 4.0 License.
164
0 1000 2000 3000 4000 5000
0
1000
2000
3000
4000
-2.5 cm
-2.0 cm
-1.5 cm
-1.0 cm
-0.5 cm
0.0 cm
0.5 cm
1.0 cm
1.5 cm
2.0 cm
2.5 cm
Figure 8. Difference in average height between the point cloud
resulting from the PPK-GNSS/IMU/LiDAR adjustment and the
point cloud resulting from the PPP-GNSS/IMU/LiDAR
adjustment. The black rectangle marks the area shown in Fig. 9.
4. DISCUSSION
In summary, the results using RTKLib, and especially the neg-
ligible difference between PPP- and PPK-derived point clouds,
are promising. The final point cloud obtained using the GNSS/-
IMU/LiDAR-integrated georeferencing approach together with
open-source RTKLib-based GNSS processing proves to be com-
parable in quality to the results of an industry-standard workflow
of GNSS/IMU integration followed by strip adjustment. With
respect to loose coupling of the GNSS measurements, we con-
clude that (1) constant or long-term shifts intrinsically cannot
be eliminated by such an adjustment and (2) time-varying errors
are correctable if sufficient redundant information from over-
lapping LiDAR strips is present. The tight-coupling of IMU
and LiDAR helps to avoid block deformations, which may oc-
cur in control-free strip adjustment. More analysis is needed to
determine the minimum amount of overlap and optimal flight
patterns to guarantee sufficient correction of the GNSS errors.
If time-varying errors can be eliminated and only a constant
datum shift is present, this could be rectified by introducing
control points as usual, but requiring a much lower number of
control points as might otherwise be necessary.
While the results are encouraging, the RTKLib software is some-
what restricted in terms of usability (parametrization of the al-
gorithms, restrictions on input data formats, possibility to use
certain precise GNSS products for PPP). Additionally, RTKLib
is not yet capable of resolving ambiguities reliably and consist-
ently in PPP, and may also have difficulties with PPK ambiguity
resolution in airborne scenarios with longer baselines. Should
this be fixed, further improvements in the direct georeferencing
accuracy may be expected in the future. The block difference
between the PPK-derived point cloud and the PPP-derived point
cloud strongly imply a systematic effect, but the difference can-
not be explained by a datum shift. A larger study area, together
with dense ground truth data, would help further investigation.
However, such data is not yet (publicly) available as data ac-
quisition requires significant time and effort, especially for the
large spatial scale of representative airborne laser scanning data
acquisitions. When acquiring reference data for the purposes
of benchmarking different georeferencing methods, special care
must be taken to ensure the accuracy of the reference data is
sufficiently high, which is challenging considering the errors
are at the centimeter-level. In data acquisitions involving mul-
tiple measurement methodologies, and carried out by multiple
parties, one must ensure the use of a single consistent datum
for all acquired data. Even then, no fully standardized methods
have been established to date for comparison of trajectories and
point clouds. Introduction of suitable benchmark data together
with standardized methods for evaluation will enable the de-
velopment of more accurate and more efficient georeferencing
methods in the future.
ACKNOWLEDGMENTS
This work was carried out as part of the project ZAP-ALS
(883660) funded by the Austrian Research Promotion Agency
(FFG2).
REFERENCES
Brun, A., Cucci, D. A. and Skaloud, J. (2022). ‘LiDAR
Point–to–Point Correspondences for Rigorous Registration of
Kinematic Scanning in Dynamic Networks’. In: ISPRS Journal
of Photogrammetry and Remote Sensing 189, pp. 185–200. doi:
10.1016/j.isprsjprs.2022.04.027.
Cadena, C., Carlone, L., Carrillo, H., Latif, Y., Scaramuzza, D.,
Neira, J., Reid, I. and Leonard, J. J. (2016). ‘Past, Present, and
Future of Simultaneous Localization And Mapping: Towards
the Robust-Perception Age’. In: IEEE Transactions on Robotics
32.6, pp. 1309–1332. doi:10.1109/TRO.2016.2624754.
Cucci, D. A., Rehak, M. and Skaloud, J. (2017). ‘Bundle Ad-
justment with Raw Inertial Observations in UAV Applications’.
In: ISPRS Journal of Photogrammetry and Remote Sensing 130,
pp. 1–12. issn: 09242716. doi:10 . 1016 / j . isprsjprs .
2017.05.008.
Emlid (2023). Emlid Studio User Documentation.url:https:
/ /emlid .com / wp- content / uploads/ 2023/ 06 /Emlid _
Studio_User_Documentation.pdf (visited on 30/06/2023).
Everett, T. (2022a). ‘3rd Place Winner: 2022 Smartphone Deci-
meter Challenge: An RTKLIB Open-Source Based Solution’.
In: 35th International Technical Meeting of the Satellite Divi-
sion of The Institute of Navigation (ION GNSS+ 2022). Denver,
Colorado, pp. 2265–2275. doi:10.33012/2022.18376.
Everett, T. (2022b). RTKLib Demo5. Version 2.4.3 b35g. url:
https:/ / github . com /rtklibexplorer/RTKLIB (visited
on 30/06/2023).
Glira, P., Pfeifer, N., Briese, C. and Ressl, C. (2015). ‘Rigorous
Strip Adjustment of Airborne Laserscanning Data Based on the
ICP Algorithm’. In: ISPRS Annals of Photogrammetry, Remote
Sensing and Spatial Information Sciences II-3/W5, pp. 73–80.
doi:10.5194/isprsannals-II-3-W5- 73-2015.
Groves, P. D. (2013). Principles of GNSS, Inertial, and Multis-
ensor Integrated Navigation Systems. 2nd ed. GNSS Technology
and Application Series. Boston: Artech House. 776 pp. isbn:
978-1-60807-005-3.
Gurturk, M. and Soycan, M. (2022). ‘Accuracy Assessment of
Kinematic PPP Versus PPK for GNSS Flights Data Processing’.
In: Survey Review 54.382, pp. 48–56. issn: 0039-6265, 1752-
2706. doi:10.1080/00396265.2020.1865016.
Jonassen, V. O., Kjørsvik, N. S. and Gjevestad, J. G. O. (2023).
‘Scalable Hybrid Adjustment of Images and LiDAR Point
Clouds’. In: ISPRS Journal of Photogrammetry and Remote
Sensing 202, pp. 652–662. doi:10 . 1016 / j . isprsjprs .
2023.07.007.
Li, T., Pei, L., Xiang, Y., Wu, Q., Xia, S., Tao, L. and Yu,
W. (2020). ‘P3-LOAM: PPP/LiDAR Loosely Coupled SLAM
with Accurate Covariance Estimation and Robust RAIM in
Urban Canyon Environment’. In: IEEE Sensors Journal.doi:
10.1109/JSEN.2020.3042968.
2Österreichische Forschungsförderungsgesellschaft, Vienna, Austria,
www.ffg.at
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLVIII-1/W3-2023
2nd GEOBENCH Workshop on Evaluation and BENCHmarking of Sensors, Systems and GEOspatial Data
in Photogrammetry and Remote Sensing, 23–24 October 2023, Krakow, Poland
This contribution has been peer-reviewed.
https://doi.org/10.5194/isprs-archives-XLVIII-1-W3-2023-161-2023 | © Author(s) 2023. CC BY 4.0 License.
165
1200 1400 1600 1800 2000 2200 2400
2800
3000
3200
3400
3600
3800
3.0
3.6
-0.3 0.8
-2.7
-1.6
-6.6
-5.3
-2.3
-1.6
-0.7
0.8
Roofs
GCPs
Exact locations
(a) PPK-based point cloud.
.
1200 1400 1600 1800 2000 2200 2400
2800
3000
3200
3400
3600
3800
3.9
4.6
0.5 1.8
-1.7
-0.5
-5.4
-4.3
-1.7
-1.3
-0.2
1.5
Roofs
GCPs
Exact locations
(b) PPP-based point cloud.
.
-10 cm
-8 cm
-6 cm
-4 cm
-2 cm
0 cm
2 cm
4 cm
6 cm
8 cm
10 cm
Figure 9. Strip differences and distances to reference targets of sub-area with high block differences, area marked in Fig. 8.
Pöppl, F., Neuner, H., Mandlburger, G. and Pfeifer, N. (2023a).
‘Integrated Trajectory Estimation for 3D Kinematic Mapping
with GNSS, INS and Imaging Sensors: A Framework and Re-
view’. In: ISPRS Journal of Photogrammetry and Remote Sens-
ing 196, pp. 287–305. issn: 09242716. doi:10 . 1016 / j .
isprsjprs.2022.12.022.
Pöppl, F., Pfennigbauer, M., Ullrich, A., Mandlburger, G., Neu-
ner, H. and Pfeifer, N. (2023b). ‘Modelling of GNSS Po-
sitioning Errors in a GNSS/INS/LiDAR-integrated Georefer-
encing’. In: Publikationen Der Deutschen Gesellschaft Für
Photogrammetrie, Fernerkundung Und Geoinformation e.V. 3.
Wissenschaftlich-Technische Jahrestagung Der DGPF, pp. 183–
196.
Rao, B. R., ed. (2013). GPS/GNSS Antennas. GNSS Technology
and Applications Series. Boston: Artech House. 404 pp. isbn:
978-1-59693-150-3.
Takasu, T. (2013). RTKLIB Ver. 2.4.2 Manual.url:http :
//www.rtklib.com/prog/manual_2.4. 2.pdf (visited on
28/08/2023).
Takasu, T. and Yasuda, A. (2009). ‘Development of the Low-
Cost RTK-GPS Receiver with an Open Source Program Package
RTKLIB’. In: International Symposium on GPS/GNSS. Jeju,
Korea, p. 7.
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLVIII-1/W3-2023
2nd GEOBENCH Workshop on Evaluation and BENCHmarking of Sensors, Systems and GEOspatial Data
in Photogrammetry and Remote Sensing, 23–24 October 2023, Krakow, Poland
This contribution has been peer-reviewed.
https://doi.org/10.5194/isprs-archives-XLVIII-1-W3-2023-161-2023 | © Author(s) 2023. CC BY 4.0 License.
166
