Quality Investigation of UAV-Based Laser Scanning with Detailed Study of Multi-Target Capability

Abstract

The use of UAV-based laser scanning systems is increasing due to the rapid development in sensor technology, especially in applications such as topographic surveys, agriculture or forestry. This paper addresses the quality analysis of direct georeferencing of a UAV-based laser scanning system and further the investigation of scan characteristics of the 2D profile scanner with multi-target detection. The precision of direct georeferencing derived from multiple repetitions of the same measurement and the use of artificial objects, such as targets, resulting in a standard deviation of < 1.2 cm for the horizontal and vertical directions. The mean absolute accuracy obtained compared to the TLS reference is < 2 cm for the horizontal direction and < 4 cm for the vertical direction. Further, different aspects focusing on the multi-target capability are investigated. The minimum distance between two planar objects for multi-target echo detection is about 1.6 m and most of the inliers are detected when the targets are more than 2.4 m apart. In addition, the relationship between the additional attribute of deviation has shown promising results for evaluating multiple echoes in terms of precision.

Full text



Quality Investigation of UAV-Based Laser Scanning
with Detailed Study of Multi-Target Capability
AnsgarDREIER,HeinerKUHLMANNandLasseKLINGBEIL
Abstract
TheuseofUAV-basedlaserscanningsystemsisincreasingduetotherapiddevelopment
insensortechnology,especiallyinapplicationssuchastopographicsurveys,agricultureor
forestry.ThispaperaddressesthequalityanalysisofdirectgeoreferencingofaUAV-based
laserscanningsystemandfurthertheinvestigationofscancharacteristicsofthe2Dprofile
scannerwithmulti-targetdetection.Theprecisionofdirectgeoreferencingderivedfrom
multiplerepetitionsofthesamemeasurementandtheuseofartificialobjects,suchastargets,
resultinginastandarddeviationof<1.2cmforthehorizontalandverticaldirections.The
meanabsoluteaccuracyobtainedcomparedtotheTLSreferenceis<2cmforthehorizontal
directionand<4cmfortheverticaldirection.Further,differentaspectsfocusingonthe
multi-targetcapabilityareinvestigated.Theminimumdistancebetweentwoplanarobjects
formulti-targetechodetectionisabout1.6mandmostoftheinliersaredetectedwhenthe
targetsaremorethan2.4mapart.Inaddition,therelationshipbetweentheadditionalattribute
ofdeviationhasshownpromisingresultsforevaluatingmultipleechoesintermsofprecision.
1Introduction
Sinceafewyears,theuseoflaserscanningonUnmannedAerialVehicles(UAV)hasin-
creasedduetotherapiddevelopmentinthecontextofsensorsforpositioningandmapping.
TheconceptofUAV-basedlaserscanningtypicallyutilizessensorsfortrajectoryestimation
withGNSS(GlobalNavigationSatelliteSystem)receiversandIMU(InertialMeasurement
Unit),aswellasasensortocapturethe3Denvironment.Sensorfusionalgorithmsareused
tocreategeoreferenced3Dpointcloudsthatareappliedinfieldssuchasmining,topographic
andbathymetricsurveying,forestry,andprecisionagriculture.(BATESetal.2021COOPSet
al.2021GUANetal.2016(6*.°3&4etal.2020MANDLBURGERetal.2020SHAH
MORADIetal.2020).
Foralltypesofapplicationswhereadditionalmodelsorparametersarederived,thequality
ofthegeoreferencedpointcloudisimportant,especiallyintermsofaccuracyandprecision.
Inthiscontext,themethodologyusedtoderivequalitymeasuresandthesourcesoferror
contributingtotheuncertaintyarenecessary.Asdiscussedlaterindetail,themainerror
sourcescontributingtotheerrorbudgetinUAV-basedlaserscanningaretrajectoryestimation,
sensorcalibration,thelaserscanneritselfandmiscellaneouserrors(SHAN&TOTH2018).
Basedonthesegroups,thelargesterroristypicallyrelatedtothetrajectoryestimation.In
$:LHVHU+UVJ,QJHQLHXUYHUPHVVXQJ +HUEHUW:LFKPDQQ9HUODJ



addition,thescanningcharacteristicsofthelaserscannerareanimportantaspectandcanalso
bealimitingfactor,dependingontheobjectsbeingscanned.Theremainingtwoerrorsources
areusuallyknownandcanbetakenintoaccountintheprocessingorcalibrationstrategies.
Forthisreason,thisstudyaimstoseparatelyinvestigatethequalityofthelaserscanneritself
andalsoincombinationwiththemulti-sensorsysteminwhichthetrajectoryestimationis
incorporated.
Consideringthequalityoflaserscanners,describedasthescancharacteristicinthefollowing,
differentqualitymeasuresarealreadyusedforthequantification.Inthisstudy,theassess-
mentapproachisbasedonaprofilelaserscannermeasuring2Dprofiles.Parametersfor
rangefinderaccuracy,resolutioncapability,rangefinderprecision,andmulti-targetcapability
canbederivedtodescribethescancharacteristics,althoughonlythelattertwoparametersare
presentedinthisstudy.Inthecontextoflaserscannersusingmultipleechoes,severalstud-
ieshaveshownthattheuseofmultipleechoesisbeneficialinderivingmodels,especially
inforestry(G
UIMARES
etal.2020).Furtherstudiesdiscussedtheimprovementofpeak
detectionalgorithmsusingthefullwaveformindetail.Missingaspectsarethesystematic
investigationofmulti-targetdetectionintermsoftheminimumdistancebetweenobjectsand
theassociatedprecisionofdetection.Theminimumdistanceforseparatingtwoobjectsis
limitedbythesensordesignandiscommonlyreferredtoasthe”verticaldeadzone”.Since
thederivedmultipleechoesareveryusefulinareaswherevegetationisrecorded,thequality
intermsoftheminimumdistancebetweenobjectsandthecorrespondingprecisionisshown
inthisstudy.Inadditiontothedetectionofmultipletargets,therangefinderprecisionofthe
laserscannerisdeterminedbasedonexistingmethods.
Apartfromtheinfluencecomingfromthelaserscanner,theadditionalerrorsoccurdueto
thetrajectoryestimationarerelevantifthegeoreferenced3Dpointcloudisused.Since
theevaluationofthetrajectoryitselfishighlycomplex,theanalysisiscommonlybased
ontheresulting3Dpointcloud.Forsuchanempiricalevaluationofgeoreferencedpoint
cloudsdifferentapproachesareused,whichcanbedividedinto(1)point-based,(2)area-
basedand(3)parameter-basedstrategies(HEINZ2021).Toassessthequalityofthedirect
georeferencing,amethodologyandmeasurementconceptarepresentedaftertheinvestigation
usingonlythelaserscanner.Thisincludestheprecisionbasedonrepetitionsofthesame
measurementflightbutalsotheaccuracyrelatedtoahigher-orderreference.Thispartofthe
studyaimstopresentareliablestrategyforqualityassessmentofthedirectgeoreferencing
ofaUAV-basedlaserscanningsystem.
Theobjectivesandresearchaspectswhicharehandledwithinthisstudyaredividedinto
oneaspectconsideringonlythelaserscannerqualityandisfollowedbyanassessmentin
combinationwiththeentiremulti-sensorsystem:
•Investigationofthescanningcharacteristicsofa2Dlaserscannertypicallyusedina
multi-sensorsystemwithoutadditionalsensorsfortrajectoryestimation.Thisincludesa
focusedinvestigationofthemulti-targetcapabilityandthepotentialforseparationbe
tweentwoobjectsintermsoftheminimumdistanceandtheprecisionoftheresulting
echoes.
•
QualityassessmentoftheentireUAV-basedlaserscanningsysteminthecontextofthe
$QVJDU'UHLHU+HLQHU.XKOPDQQDQG/DVVH.OLQJEHLO

urban environment. Additional focus is on the precision due to repeated measurements, but
also on the accuracy of the georeferenced point cloud. Moreover, the link between
investigations performed with laser scanner-only experiments and the use of the multi-
sensor system is established.
The paper is organized as follows: Section 2 presents the methodology of UAV-based laser
scanning and multi-target capability. This is followed by Section 3 describing the measure-
ment concept and approach for quality analysis. Section 4 presents the results and Section 5
concludes the paper.
2 Methodology of UAV-based laser scanning and multi-target
capability
The basic principle and in particular the errors in the processing of UAV-based laser scanning
measurements are presented in the following. Typically, a combination of sensors is used
for trajectory estimation with the position and orientation of the UAV and the laser scanner.
The processing is shown in Figure 1, starting with GNSS baseline processing using raw data
from a receiver on the UAV and an additional master station. Afterwards, these are fused
Fig. 1: Processing of UAV-based laser scanning measurements resulting in the direct geo-
referenced 3D point cloud
with IMU measurements to derive the entire trajectory of the measurement campaign. In
the next step, the time-synchronized laser scan data is georeferenced using the trajectory
and calibration parameters, resulting in a georeferenced 3D point cloud. In addition to the
presented processing, flight strip adjustment or trajectory optimization is performed by using
flight strips that have covered the same area. The explained processing can also be formulated
with the equation
⎡
⎢
⎢
⎣
xe
ye
ze
⎤
⎥
⎥
⎦
=⎡
⎢
⎢
⎣
tx
ty
tz
⎤
⎥
⎥
⎦
+Re
n(L,B)Rn
b(φ,θ,ψ)·⎡
⎢
⎢
⎣⎡
⎢
⎢
⎣
Δx
Δy
Δz
⎤
⎥
⎥
⎦
+Rb
s(α,β,γ)·⎡
⎢
⎢
⎣
0
d·sinb
d·cosb
⎤
⎥
⎥
⎦⎤
⎥
⎥
⎦
(1)
4XDOLW\,QYHVWLJDWLRQRI8$9%DVHG/DVHU6FDQQLQJ 

forthederivationofeach3Dpoint[xe,ye,ze]Tincludingtheparametersforthetrajectory,sen-
sorcalibrationandlaserscannermeasurements(VOSSELMANN&MAAS2010).Thesensor
calibrationdescribesthespatialrelationbetweenthe2DlaserscannerandtheGNSS/IMU
unitusingaleverarm[Δx,Δy,Δz]Tandthethreeboresightanglesα,βandγ.Further,the2D
laserscannermeasurementswith[0,d·sinb,d·cosb]Tarebasedonthemeasuredranged
andthescanangleb.Trajectoryparametersaretheorientationangleswithrollφ,pitchθand
yawψ,itspositionwithtranslationvector[tx,ty,tz]TandellipsoidallongitudeLandlatitudeB.
AdditionalrotationmatricesRdescribetherotationbetweenthesensor(s),body(b),
navigation(n)andearth(e)-frame.
Forthequalityanalysisofthemulti-sensorsystemandthusthe3Dpointcloud,theerrors
areimportant,whichareacombinationofseveralsystematicandrandomerrors(SHAN
&TOTH2018).ThemainsourcesoferrorassociatedwithUAV-basedlaserscanningare
trajectoryestimation,systemcalibration,thelaserscanneritself,andmiscellaneouserrors
suchaswrongtimesynchronization(BALTSAVIAS1999).Adetailedlistoferrorscanbe
foundinTable1forthementionedgroupsoferrorsources.InthecontextofUAV-based
laserscanning,thelargesterrorstypicallyoriginatefromtrajectoryestimation,astheseerrors
arealsosystematicallypropagatedinerrorsofthegeoreferenced3Dpointcloud.Errors
Table1:MajorerrorsourcesforUAV-basedlaserscanningsystems(refertoSHAN2018).
Errors
Trajectory estimation errors in position and orientation of the sensor platform
System calibration
error in lever arm (GNSS antenna and IMU)
error in lever arm (laser scanner and IMU)
boresight angle error between IMU body and laser scanner
Laser scanner
range measurement error
object characteristics
atmospheric refraction
Miscellaneous errors time synchronisation
sensor mounting rigidity
due to system calibration also introduce systematic errors, but typically these calibration
parameters can be estimated in terms of lever arms and boresight angles as part of processing
data sets. A further option is the use of specific flight patterns with the subsequent estimation
of calibration parameters. Therefore, the trajectory estimation and the laser scanner have the
greatest influence on the resulting quality of the point cloud, and their influence is investigated
in the study. Since the separation of the sources of error in the use of the derived point cloud
is very complex, the laser scanner is first studied separately and then in combination with the
multi-sensor system.
The laser scanners used in UAV-based laser scanning are available with different scanning
$QVJDU'UHLHU+HLQHU.XKOPDQQDQG/DVVH.OLQJEHLO

mechanisms, specifications, and also scanning patterns (SHAN &TOTH 2018). In addition
Fig. 2:
.VMUJUBSHFU EFUFDUJPO CBTFE on
full waveform processing
to these different features, the capability of multi-target acquisition is commonly incorporated
in laser scanners using time of flight t echnology. This capability uses the full w aveform of
the signal to detect multiple objects within the laser beam, as shown in Figure 2.
Peak detection algorithms use the full waveform of the reflected signal recorded by the re-
ceiver to determine corresponding distances for multiple targets hit by the laser beam. This
capability is particularly used in the context of forestry or more generally vegetation, where
the detected echoes separate the canopy and the ground (GUIMAR ˜
AES et al. 2020). The multi-
target capability depends on several aspects, such as the footprint of the laser beam, the
reflectivity and the size of the objects, but especially the distance between the targets (SHAN
&TOTH 2018).
3 Data acquisition and approach for quality assessment
The data acquisition and corresponding approach for quality assessment are discussed for
the only-laser scanner analysis in Section 3.1 followed by the use of the entire multi-sensor
system in Section 3.2. Since the UAV-based laser scanning system shown in Figure 3 is used
for the study, the components included are explained as follows. The platform consists of
the DJI Matrice 600 Pro in combination with the laser scanning system RIEGL miniVUX-
SYS. The latter one is further a combination of the integrated IMU/GNSS sensor Trimble
APX-20 UAV, which is used for trajectory estimation in post-processing. In addition, the
RIEGL miniVUX-2UAV is integrated as the 2D profile l aser s canner and the RGB camera
Sony Alpha 7R. Important specifications are the given error budget for trajectory estimation
with <0.10 m standard deviation for the position in vertical and <0.05 m for the position
in horizontal direction. The orientation is given with 0.015 ◦for roll & pitch and 0.035 ◦
for heading. Furthermore, the 2D laser scanner uses a online waveform processing for multi-
target detection (up to 5 targets) and is given with 15 mm accuracy and 10 mm precision
specified for the range of 50 m. Important specifications are further the laser beam divergence
4XDOLW\,QYHVWLJDWLRQRI8$9%DVHG/DVHU6FDQQLQJ 

(FWHM) with 1.6 x 0.5 mrad. In the following, the measurement concept and approach
for evaluation are given for the 2D laser scanner followed by the explanation using the
UAV-based laser scanning system.
3.1 Scan characteristics and multi-target capability of a 2D laser scanner
Within the scope of this study, the quality measures describing the scan characteristics based
on range precision and precision of multi-target detection are investigated. For the first cri-
terion, the derivation of an intensity-based stochastic model of the laser scanner has been
developed in several studies. Corresponding methodologies have been shown for 3D scan-
ners (SCHMITZ et al. 2019WUJANZ et al. 2018WINIWARTER et al. 2020) as well as for 2D
profile scanners (HEINZ et al. 2018). The range precision of laser scanners is affected by
atmospheric and environmental conditions, instrument mechanics, object properties, and
scan geometry. Therefore, the range precision is first derived from multiple scans with vari-
ations for the mentioned aspects and then transformed into a stochastic model. Individual
range precision can be estimated from multiple profiles of a given target, estimating a line GJU
BOEUIF DPSSFTQPOEJOHSFTJEVBMTPGUIFMJOFGJU5IFTFSFTJEVBMTBSFVTFEUPDBMDVMBUFƐSBOE
UIF DPSSFTQPOEJOH JOUFOTJUZ WBMVF SFQSFTFOUFE CZ UIF NFBO BNQMJUVEF À8
6+"/; FU BM
5IFOUIFJOUFOTJUZCBTFETUPDIBTUJDNPEFMJTEFSJWFECZ
σr=a·¯
Ab+c(2)
using multiple scans with a change in scan geometry, scan parameters, and object properties.
The unknown parameters a, b and c are estimated in the adjustment.
The second aspect considered when using the laser scanner only is the quality of the echoes
obtained by detecting multiple targets. This ability depends on several aspects such as the
footprint of the laser beam, the reflectivity and the size of the objects within the laser beam,
but also the distance between the objects (SHAN &TOTH 2018). The methodology for
evaluation is explained using the measurement approach shown in Figure 4, where the laser
scanner mounted on a pillar measures two targets separated by the variable distance d. Since
the scan profile passes over both targets, it includes laser beams hitting both targets and thus
the possibility of detecting multiple echoes. In the analysis, the single echoes (reflection
only from one target) on the first and second Uargets are used for a line fit that defines the
$QVJDU'UHLHU+HLQHU.XKOPDQQDQG/DVVH.OLQJEHLO
Fig. 3:
5IFVTFE6"7CBTFEMBTFSTDBOOJOH
TZTUFN FRVJQQFE XJUI 3*&(- NJOJ
7694:4BOE4POZ"MQIB3

Fig. 4: Measurement concept for the investigation of multi-target capability using two tar-
gets with a variation between both defined by distance d
individual target. Then, the residuals of the multiple echoes to this line fit are used to separate
outliers and inliers based on a defined threshold. With this approach, multiple echoes can be
evaluated not only according to their occurrence but also according to their corresponding
precision. The measurement approach in this study is set up as shown in Figure 4, with a
fixed distance for the second target and varying the distance between the two targets. This is
studied for different materials, but mainly for varying the distance to the second target with
20, 40, 60, 80, and 100 m outdoors and 22.5 m indoors. For the individual scan setup, it is
further possible to estimate the number of laser beams illuminating both targets per profile
based on the geometry and thus a theoretical number for multiple echoes.
3.2 Assessment of direct georeferencing of a UAV-based laser scanning system
The evaluation of direct georeferencing with the entire UAV-based laser scanning system
based on the georeferenced point cloud is provided below. For this purpose, four identical
flights, as shown in Figure 5, are performed using the UgCS flight planning software. The
flight parameters are a fMight height of 25 m, a flight speed of 3 m/s, and a laser scanner
line speed of 58.92 lps. With this configuration, the average point spacing on the ground is
0.055 m and the point density is about 333 pts/m2. In addition to the UAV measurements,
TLS measurements are performed to generate a reference specifically for the targets shown in
Figure 5. Since direct georeferencing will be evaluated, the TLS point cloud is georeferenced
using a control network, with accuracy reported as <1cm(D
REIER et al. 2021). The
dataset is used to derive quality measures that account for the accuracy, precision, and noise
of the georeferenced point cloud. Therefore, planar surfaces are utilized for plane estimation
and subsequent quantification of noise based on residuals. Moreover, the included targets are
used, on the one hand, for the evaluation of precision based on repeated measurements of
the same area. On the other hand, the TLS reference is used for evaluating the accuracy of
direct georeferencing. Since additional optimization strategies, as integrated into the RIEGL
software RiPRECISION, are used, alignment errors between flight strips are detected by the
performed evaluation strategies.
4XDOLW\,QYHVWLJDWLRQRI8$9%DVHG/DVHU6FDQQLQJ 

Fig. 5: Study area including multiple strip measurements performed with the UAV and
ground targets used within the evaluation. https://www.google.de/maps/ (accessed
on 14 May 2021)
4 Results
The following results are structured with a detailed discussion of the laser scanner study and
the quality aspects of range precision and multi-target capability. In addition, the results are
partially applied to a case study in an urban environment where the additional aspects of
accuracy and precision using the UAV-based laser scanning system are discussed.
4.1 Results of the scan characteristics using a 2D laser scanner
The intensity-based stochastic model for the RIEGL miniVUX-2UAV presented in 3.1 is
shown in Figure 6 with the relationship between amplitudes and range precision. In addition
to the functional relation σr(¯
A)estimated using a Gauss-Helmert model, the results mea-
sured using different materials, incidence angles and scan parameters are plotted. Distances
between 20 m and 100 m are included in the various setups. In general, the relationship fits
better for higher amplitude measurements and less well for the lowest amplitude values, so
measurements with very low amplitudes (<15 dB) are excluded from the model. As can be
seen in Figure 6, the functional relationship represents the stochastic model well and can be
used to calculate range precision if the amplitudes are known.
Aside from the stochastic model, the multi-target capability is explained in more detail in
terms of the minimum distance between the objects and the corresponding precision of the
multiple echoes. The results presented are based on the measurement setups already ex-
plained and for distances between 20 m and 100 m for the second target. The distance be-
$QVJDU'UHLHU+HLQHU.XKOPDQQDQG/DVVH.OLQJEHLO

Fig. 6: Intensity-based stochastic model based on amplitude and corresponding range pre-
cisionforRIEGLminiVUX-2UAV
tween the two targets is between 1.5 m and 5 m, which is based on previous experiments.
Figure 7 shows the residuals for multiple echoes for wooden targets at distances of 1.8 m,
2.1 m, 2.4 m, and 3.0 m. The second target is placed at a distance of about 20 m and the
detected echoes are measured using 100 profiles of the laser scanner. The scanning param-
eters used are the pulse repetition rate of 200 kHz and 10 lps. Moreover, a threshold based
Fig. 7: Residuals of multiple echoes for wooden targets and four different distances between
the first (blue) and second (orange) target (1.8 m, 2.1 m, 2.4 m and 3.0 m).
on the manufacturer’s specified precision σR(10 mm) is used to evaluate the precision of the
detected echoes. In this case, the manufacturer’s specification is applied because the stochas-
tic model does not clearly agree with measurements resulting from the detection of multiple
targets. Several conclusions can be drawn from this initial plot. First, the number of detected
echoes increases as the distance between targets increases. Also, the residuals for the first
target are always smaller than for the second target (e.g. 2.1 m distance). In addition, the
residuals are quite large (up to 0.5 m) for the shortest distances and improve as the distance
4XDOLW\,QYHVWLJDWLRQRI8$9%DVHG/DVHU6FDQQLQJ 

between targets increases. The number of detected multiple echoes per profile (not shown in
the figure) reaches a maximum of about 7 echoes, which is even higher than calculated based
on the geometry. This slight difference could be due to a simplified model of the laser beam
propagation.
The systematic effect between targets can also be shown for the whole experiment outside
with distances up to 100 m and scan parameters with 200 kHz pulse repetition rate and 50 lps.
Figure 8 shows the percentage of classified inliers (3σR) as a function of the distance between
targets with separation for echoes on the first (T1) and second (T2) targets. The first inlier for
Fig. 8: Percentage of multiple echoes classified as inliers within 3σRfor distances up to
100m
the first target is observed just below 2 m distance between targets, followed by a rapid in-
crease to 2.4 m with more than 95%. As shown previously for the residuals, the percentage of
inliers for the second target is much lower at the shortest distances and increases more slowly
compared to the first target. In general, a systematic separation between the two targets can
be seen, which might be less noisy if the measurement is carried out indoors under more con-
trolled conditions. Only small differences are seen between the results for distances between
20 and 100 m, showing that multiple target detection works well at increasing distances.
Apart from the general investigation of the multi-target capability, the dependence on ad-
ditional attributes of the laser scanner is also considered. One attribute derived for each
measurement is the deviation value as a measure of the trustworthiness of a point. It is de-
fined as the difference of the echo pulse shape from the pulse shape received from a plane
target at a perpendicular angle of incidence (so-called system response) and is given with-
out unit. The smaller the deviation, the better the reflection characteristics and the possible
quality. Based on this idea, Figure 9 shows the mean deviation values corresponding to the
detection of multiple targets for the same experiment and an additional data set containing
different materials. In this relationship, a systematic decrease of the deviation values with
increasing distance between the targets can be seen. In particular, very short distances with
high deviation values indicate less accurate measurements, which was also shown in Figure
8. Additionally, the difference in deviation values is present at different distances, which
is quite an unexpected behavior. The presented results establish a relationship between the
deviation value and the precision of multi-target detections, which can be used to evaluate
$QVJDU'UHLHU+HLQHU.XKOPDQQDQG/DVVH.OLQJEHLO

Fig. 9: Mean deviation of multiple echoes related to the distance between the target for
differentmaterialsanddistances(secondtarget)
derived points in applications such as forestry or urban environments. It must be empha-
sized that the measurement setup with two planar objects simplifies the interaction between
the laser beam and objects and cannot necessarily be transferred to applications where the
objects have completely different shapes.
4.2 Results for quality assessment using the UAV-based laser scanning system
In addition to the specific aspects of the scan characteristics related to range precision and
multi-target capability, the effects of errors in direct georeferencing and hence trajectory are
also discussed. When processing the four data sets acquired with the cross-flight pattern
shown in Figure 5, a virtual reference station is used for GNSS processing and additional
trajectory optimization within RIEGL’s RiPRECION. This additional optimization step im-
proves the point cloud and, in particular, reduces alignment errors caused by the uncertainty
of the trajectory estimation. In the first step, the 12 targets within the study area are used
to quantify the noise based on the residuals of a plane fitting. The distribution of estimated
noise values for 12 targets and four flight repetitions has no outliers and a mean noise value of
6 mm, which fits the intensity-based stochastic model derived without the additional sensor
system. This shows a comprehensive result from both studies and good performance using
additional optimization strategies.
In addition to the noise, a point-based evaluation is performed in comparison to the TLS
reference using the estimated target center coordinates. Figure 10 shows the differences of
all target centers from the TLS mean, indicated by dx,dy, and dz (lighter colors). Since
this histogram already contains the repetitions of the same flight, these values indicate the
precision of the georeferenced point cloud measured with the sensor system. The precision
described by the standard deviation is given as 0.8 cm for the east, 0.5 cm for the north
and 2.1 cm for the height component. Moreover, the histogram shows the differences from
the reference indicated with the mean values δx,δyand δzand the individual differences
(darker colors). The mean differences obtained give -1.1 cm for east, 1.1 cm for north, and
4.3 cm for height direction with corresponding RMSE of 1.8, 1.0 cm, and 2.2 cm. This
combination describes the accuracy of direct georeferencing. The systematic offset in the
vertical direction of about 4 cm could be due to system calibration but can be corrected in
4XDOLW\,QYHVWLJDWLRQRI8$9%DVHG/DVHU6FDQQLQJ 

Fig. 10: Differences of estimated target coordinates in comparison with TLS reference
(darker colors). The comparison to the mean target coordinates (lighter color). The
results are based on 12 different targets and four repetitions of the same flight. Index
j describes the individual target and index i the corresponding measurement flight.
practice if necessary. Therefore, a single ground control point (GCP) with accurate height
information can be included in the processing. Overall, the differences in horizontal and
vertical are small with regard to the manufacturer’s specifications. This evaluation is used to
illustrate the effects of direct georeferencing on a specific case study. Since the effects of the
precision of the laser scanner are known, the errors in the trajectory estimation can be better
evaluated.
5 Conclusion
In this paper, the quality assessment of the scan characteristics of a 2D profile laser scan-
ner and in combination with the UAV-based laser scanning system is presented. In the first
part, the derivation of an intensity-based stochastic model to describe the range precision has
shown a good agreement with the included observations. Furthermore, the methodology for
analyzing the multi-target capability in combination with the RIEGL miniVUX-2UAV was
presented. The minimum distance between two planar objects for multi-target echo detection
is about 1.6 m and most of the inliers are detected when the targets are more than 2.4 m
apart. In addition, the relationship between the additional attribute of deviation has shown
promising results for evaluating multiple echoes in terms of precision. The precision and ac-
curacy investigated using the UAV-based laser scanning system represent a quantification of
the errors included by direct georeferencing, since the capability of the laser scanner is well
$QVJDU'UHLHU+HLQHU.XKOPDQQDQG/DVVH.OLQJEHLO

known. In addition to the results presented, a variety of further aspects can be identified:
•
Sincethemeasurementconceptofmulti-targetdetectionusesonlyplanarobjects,
transferringtheresultstomorecomplexobjectsisquitedifficult.Futureworkwill
includemorecomplexobjectsfortheevaluationofmulti-targetcapability.
•Theassessmentofmultipleechoesusingthedeviationattributehasbeenshownwiththe
experimentalsetup.Thetransfertoarealcaseliketheapplicationinforestryseemsrea
sonable.
•Theevaluationstrategyofdirectgeoreferencingiscrucialiftheerrorsaffectingqualityare
tobeseparated.Inparticular,theeffectsoftrajectoryestimationwouldbeclearerifthe
trajectorycouldbeevaluatedseparately,asitisthecasewiththelaserscanner.
'VOEJOH:
ThisworkwasfundedbytheDeutscheForschungsgemeinschaft(DFG,German

ResearchFoundation)underGermanysExcellenceStrategyEXC2070390732324.
#JCMJPHSBQIZ
BALTSAVIAS,E.P(1999):Airbornelaserscanning:basicrelationsandformulas.In:ISPRS
Journalofphotogrammetryandremotesensing54(2-3),199214.
B
ATES
,J.,M
ONTZKA
,C.,S
CHMIDT
,M.J
ONARD
,F.(2021):fUASLiDARCropLAI
EstimationsfromCanopyDensity.In:2021IEEEInternationalGeoscienceandRemote
SensingSymposiumIGARSS,7720-7723.
C
OOPS
,N.C T
OMPALSKI
,P.,G
OODBODY
,T.RH,Q
UEINNEC
,M.,L
UTHER
,J.E,
BOLTON,D.K,WHITE,J.C,WULDER,M.A,VANLIER,O.R  HERMOSILLA,T.
(2021):Modellinglidar-derivedestimatesofforestattributesoverspaceandtime:Are-
viewofapproachesandfuturetrends.In:RemoteSensingofEnvironment260,112477.
%
3&*&3
 " +
"/&/
 + ,
6)-."//
 )  ,
-*/(#&*-
 -2VBMJUZ"OBMZTJTPG
(6"/ ) -* + $"0 4  :6 :6TFPGNPCJMF-J%"3JOSPBEJOGPSNBUJPO
(6*."3&4 / 1%6" - ."326&4 1 4*-7" / 1&3&4 &  4064" + + 
)&*/;& #FJUShBHF[VS,BMJCSJFSVOHVOE&WBMVJFSVOHWPO.VMUJTFOTPSTZTUFNFOGhVS
)&*/; & .&55&/-&*5&3 . ,6)-."// )  )0-45 $  4USBUFHZ GPSEFUFS
."/%-#63(&3 ( 1'&//*(#"6&3 . 4$)8"3; 3 '-3: 4  /644#"6.&3 -
(2020):ConceptandPerformanceEvaluationofaNovelUAV-BorneTopo-Bathyme
tricLiDARSensor.In:RemoteSensing12(6),986.
4XDOLW\,QYHVWLJDWLRQRI8$9%DVHG/DVHU6FDQQLQJ 
%JSFDU(FPSFGFSFODJOHJO"TQFDUTPG"CTPMVUF"DDVSBDZBOE1SFDJTJPOGPSB6"7#BTFE
-BTFS4DBOOJOH4ZTUFN*O3FNPUF4FOTJOH
JOWFOUPSZ"SFWJFX*O*OUFSOBUJPOBM+PVSOBMPG*NBHFBOE%BUB'VTJPOȰ
'PSFTUSZSFNPUFTFOTJOHGSPNVONBOOFEBFSJBMWFIJDMFT"SFWJFXGPDVTJOHPOUIFEBUB
QSPDFTTJOHBOEQPUFOUJBMJUJFT*O3FNPUF4FOTJOH
LJOFNBUJTDIFT-BTFSTDBOOJOH1I%5IFTJT6OJWFSTJUZPG#POO#POO
NJOJOHUIFTUPDIBTUJDEJTUBODFDIBSBDUFSJTUJDTPGUIF%MBTFSTDBOOFS;'QSP˳MFSB
XJUITQFDJBMGPDVTPOUIFDMPTFSBOHF*O4FOTPST

SCHMITZ,B.,HOLST,C.,MEDIC,T.,LICHTI,D.D.KUHLMANN,H.(2019):Howto
SHAHMORADI,J.,TALEBI,E.,ROGHANCHI,P.HASSANALIAN,M.(2020):ACompre-
hensiveReviewofApplicationsofDroneTechnologyintheMiningIndustry.In:Drones
4(3),34.
SHAN,J. TOTH,C.K(2018):Topographiclaserrangingandscanning:principlesand
V
OSSELMAN
,G.

M
AAS
,H.-G.(2010):Airborneandterrestriallaserscanning. $3$QSFTT
WUJANZ,D.,BURGER,M.,TSCHIRSCHWITZ,F.,NIETZSCHMANN,T.,NEITZEL,F.
WINIWARTER,L.,ANDERS,K.,WUJANZ,D.H½FLE,B.(2020):Influenceofrangingun-
DFSUBJOUZ PG UFSSFTUSJBM MBTFS TDBOOJOH PO DIBOHF EFUFDUJPO JO UPQPHSBQIJD % QPJOU
DMPVET*O*4134"OOBMTPGUIF1IPUPHSBNNFUSZ3FNPUF4FOTJOHBOE4QBUJBM*OGPSNB
UJPO4DJFODFT
$QVJDU'UHLHU+HLQHU.XKOPDQQDQG/DVVH.OLQJEHLO
&GmDJFOUMZ%FUFSNJOFUIF3BOHF1SFDJTJPOPG%5FSSFTUSJBM-BTFS4DBOOFST*O4FOTPST

QSPDFTTJOHT$3$QSFTT3PVUMFEHF64"
3PVUMFEHF64"
,&345&/51%FUFSNJOBUJPOPG*OUFOTJUZ#BTFE4UPDIBTUJD.PEFMTGPS5FSSFT
USJBM-BTFS4DBOOFST6UJMJTJOH%1PJOU$MPVET*O4FOTPST