OBJECT ORIENTED SOFTWARE ENGINEERING FOR DESIGNING AN AERIAL SURVEY LIDAR SIMULATOR

Rakesh Kumar Mishra, Bharat Lohani

Geoinformatics division, Indian Institute of Technology Kanpur, Kanpur 208016, INDIA – rkmishra@iitk.ac.in

KEY WORDS: Object-oriented, Software engineering, LiDAR, Modelling of sensor, Research and education tool ABSTRACT: This paper describes the object oriented design methods used to develop a software system to simulate the functioning of an aerial survey LiDAR instrument. Object-oriented design is a design strategy where system designers think in terms of ‘things’ instead of operations or functions. The goal of this paper is to model functions of an actual LiDAR instrument by using object oriented software engineering approach so that the developed software could be maintainable, reliable and scalable. The simulator is conceived with three components, namely terrain component, sensor component and trajectory component. Each component has been divided into their sub modules and designed independently. The terrain component deals with generation of bare terrain and objects on top of the terrain. The sensor component deals with design of commercially available sensors or a generic sensor. And the trajectory component deals with modelling of platform parameters, viz., velocity, roll, pitch, yaw and accelerations. Numerical methods are used to solve complex problems for generating LiDAR data from simulated terrain and flights. LiDAR data files are generally very large. Considering this, special data structures and file formats have been designed to improve the performance and to solve the memory problems. This user friendly GUI based simulator, developed in JAVA programming language, is an ideal tool for research and education.

1. INTRODUCTION

1.1 Background

The last decade has seen manifold growth in the use of aerial survey LiDAR (Light Detection and Ranging) technology. Due to the main advantage of measuring topography through highly dense and accurate data points which are captured at high speed, the LiDAR technology has found several interesting applications (Lohani, 2001; Queija, et al., 2005). 1.2 Object-oriented software engineering

Software engineering has traditionally been an expensive and time-intensive process. Object-oriented analysis and design is the principal industry-proven methodology that answers the call for a more cost-effective, faster way to develop software and systems. Object-oriented technology cuts development time and overhead, leading to faster time to market and significant competitive advantage, enabling software engineers to produce more flexible and easily maintainable applications. 1.3 Characteristics of LiDAR simulator

A LiDAR simulator is aimed at faithfully emulating the LiDAR data capture process with the use of mathematical models under a computational environment. Basically, data generated by simulator should exhibit all characteristics of data acquired by an actual LiDAR sensor. Literature reveals that only a few attempts have been made by researchers to develop simulator for LiDAR instrument. These efforts are limited in their scope as either these consider the effect of only single parameter on one kind of object (Holmgren et al., 2003) or inaccurate scanning pattern (Beinat and Crosilla, 2002). Another attempt is made (Kukko and Hyyppa, 2007) using MATLAB however the simulator has limitations as it is designed only for test purpose and does not offer flexibility and completeness. More focused and comprehensive efforts have been made to simulate the return waveform from a footprint (Sun and Ranson, 2000; Tulldahl and Steinvall, 1999).

1.4 Requirement of LiDAR simulator software

Considering the significance of LiDAR technology there is a need to introduce LiDAR technology to students at undergraduate and postgraduate level. LiDAR instrument and data are costly. Collecting LiDAR data with varied specifications, as are desired for classroom teaching and laboratory exercises, may not be viable considering the cost and availability of instrument. A simulator can generate various kinds of data, as and when desired, at minimal or no cost. This data could be very useful for conducting laboratory exercises. User control over the entire data generation process in simulator can also help students in understanding the functioning and limitation of LiDAR instrument. Further, error sources and their effect on LiDAR data can be understood. In any laboratory exercise availability of ground truth is fundamental. While it may be difficult and expensive to collect ground truth in case of actual LiDAR data, for simulated data the ground truth is readily and accurately available. In addition to education, there are several other applications where data generated by simulator can be employed. In particular, for testing the information extraction algorithms for their performance over a wide variety of data is conveniently possible with simulated data. It can also be used for LiDAR flight planning.

2. BENCHMARK FOR LIDAR SIMULATOR

Considering the aforesaid applications the following benchmarks are set for the simulator:

Simulator should employ a user-friendly GUI (Graphical User Interface.) Simulator should be easily extendable and maintainable. Simulator should be designed for wider distribution over various computational platforms.

3

IpTsnftci

Eaiomd 3

TMaouges

The simulsystem whifriendly muIt should widely avaiThe simulasensor paraof introducLiDAR. Simulator slike surfaceThe outputLiDAR for

3.

3.1 Software c

In view of programming (OThis technologysoftware. OOP number of entifunctions arounthe code reusabicomponents in ain Figure 1.

Figure 1. Basi

Each componenaccordingly. Eaintegration geneof these componmathematical mdevelopment of

3.2 Terrain co

Terrain componMathematical suare designed to objects. Each musing GUI. The generation. Furteach module toshown in Figure

Mathemati

Terrain componen

ator should coich can explain ultimedia techniqsimulate a genilable sensors inator should facilameters as in accing errors in

should facilitate es. t data should brmats.

SOFTWARE

components

the aforesaid OOP) has been uy offers a new a

allows decomities called objed these objects.ility. The simuaddition to inpu

ic components a

nt is divided intoach module takeserates LiDAR danents is realised

models. The folthe present vers

omponent

nent (Figure 2) isurface, Raster a

simulate bare emodule has facil

marked area bether, the user haso create the terre 3.

Figure 2. Terr

Terrain com

ical surface

INP

Sencompnt

OUT

me along withconcepts of LiD

ques. neric LiDAR sen market. litate selection o

ctual case along various compon

data generation

be available in

DEVELOPME

requirements used for developiand powerful m

mposition of a ects and then A major advan

ulator is composut and output mo

and their integrat their sub modus input from the ata as desired byd by developingllowing paragrapion of software.

s combination ond fractal. Thesearth surface anlity to mark arecomes availables option to chanrain of his choi

rain component

mponent

Raster F

PUT

nsor ponent

Tco

TPUT

h a help/tutoriaDAR using user

ensor and some

of trajectory andwith the facility

nent systems o

for actual earth

commonly used

ENT

object-orienteding the simulato

model for writingproblem into abuilds data andntage of OOP ied of three basicodules as shown

tion in software

ules and designeduser, while thei

y the user. Eachg algorithms andphs describe the

of three modulesse three modulend above groundea of interest bye for LiDAR datage parameters oice. The GUI i

Fractal

Trajectory omponent

al r-

e

d y f

h-

d

d r. g a d s c n

d ir h d e

s, s d y a f s

3.2.1 MIn this, tequations, selection oare AX+BYZ=A [sinZ=A [sin

3.2.2 RIn this, a different oGUI permobjects onobjects. Thfile formaaddition, itLiDAR da 3.2.3 FrHere, a nalgorithmschosen aregenerated 3.3 Senso

Two comma generic modules. Dof the abofor LiDARfiring freqparametersspecificatirange of (Lohani a

Mathematical surerrain is modewhich yield ear

of these surfaces

Y+CZ+D=0 n(X/B) - sin (n(X/B) – cos

Figure 3. GU

Raster terrain using th

objects over the mits to choose thn it. Further, it phe generated rasat to improve tt is possible to i

ata can be genera

ractal natural-looking ts. This moduleea by using GUby varying inpu

or component

mercially availabsensor (Figur

Developed GUI ve sensors and sR data generatio

quency, type of s is constrainedons. The generparameters. Fo

and Mishra,2007

Figure 4

Senso

Generic senso

face eled by using rth like surfacess and their param

XY / C)] (XY / C)]

UI for terrain com

he above mathesurfaces are rast

he surface type permits to drag ster data is savedthe performanceimport an existinated.

terrain is genere generates ranUI. Different typut parameters.

ble sensors (ALre 4) are desig(Figure 5) faciliset their parameon, viz., scan ascanning (Figur

d in commerciaric sensor perm

or algorithmic d).

4. Sensor compo

r component

r ALTM

mathematical s. The software pmeters. A few o

(1)

mponent.

ematical surfaceterized. The devand different tyand resize the

d in specially dee of the softwang raster file for

rated by using ndom fractal fopes of fractals c

LTM and ALS 5gned as indepitates selection o

eters which are dangle, scan freqre 6) etc. The ral sensors as pe

mits selection odetail please re

nent

M ALS 50

surface permits f these

)

es and veloped ypes of placed

esigned are. In which

fractal for the can be

0) and endent of any desired quency, ange of r their

of any efer to

3

Taeasogaa

3As(sma

Fi

Figure

3.4 Trajectory

This componenattitude. Locatioeach instance oacceleration. Ansemi random (coof altitude angenerates trajectat each instanceand Mishra-2007

3.4.1 LocatioA trajectory (resingle flight) is (point of originsystem at each modules, velocacceleration mo

T

Lo

Acceleration

gure 5. GUI for

e 6. Sinusoidal an

y component

nt is divided inon module giveon the basis of nd the altitude montrolled) Roll, nd location mtory coordinate

e. For algorithm7).

Figure 7. Trajec

on eferred as fligh

defined by then of laser vectinstance of firi

city and accedule generates p

Trajectory com

ocation

Velocit

Sensor compone

nd zig-zag scan

nto two modulees the position f its sub modulemodule is respon

Pitch, and Yawmodules, traject

(GPS coordinatmic detail please

ctory component

ht line hencefore location of lastor) in the absoing of laser pulseleration are dpseudo-random a

mponent

Attitude

ty

ent

pattern

es, location andof aeroplane a

es, velocity andnsible to generate. The integrationtory componentes) of aeroplanerefer to (Lohan

t

rth to indicate aser mirror centeolute coordinatese. The separatedeveloped. Theacceleration in x

d at d e n

nt e

ni

a r e e e

x,

y and z controlled Velocity mdirection amodule. Invelocity ivelocity, lo

3.4.2 ADue to intcertain limattitude mrandom attchanging tbe set to zbe importe

3.5 Integ

The compof the laseThe point Dependingdirection othe optimucase, the fdirection oto (Lohani

Figure 8. A(arr

direction withiby changing the

module calculateat each instance n software, it is a.e. without accocation of the ae

Attitude ternal and extern

mits and may exhmodel is designtitude. The attituthe parameters gzero. Further, ated form file.

Figure 7. GUI

gration of comp

ponents discusseer vector and itof intersection y

g the area choof flight lines (Fium direction is flight direction of the chosen ari and Mishra, 20

Area of interest (rows) and swath

in the permissie parameters giv

es velocity of aebased on the inp

also possible to deleration variat

eroplane can be d

nal forces, the ahibit random behned. This moduude can be changgiven in the softwttitude values of

I for trajectory c

onents

ed above are intts intersection wyields the coordin

sen for LiDARigure 8) is eitherdetermined by is determined b

rea. For algorithm07).

(gray), flight lin

hs with overlap (t

ible limit. It cven in the softwa

eroplane in x, yput of the acceledo flight with cotion. On the badetermined.

attitude changes haviour. To realiule generates pged within the liware. Attitude caf actual flight ca

omponent

tegrated by genewith simulated tnate of terrain p

R data generatior chosen by the uthe software. In

by using the primic detail pleas

es in optimal dirthin rectangles)

can be are.

y and z eration onstant asis of

within ize this pseudo imit by an also an also

eration terrain. oint.

on the user or n later incipal e refer

rection

TaTsc Tffrgi TvarTat TrfaTkt 3

Ldabpviie

Wwcfecks 3

Tod

TTV

Li

The number of and sensor paramThe initial and software. The tchosen for LiDA

Trajectory compflight line depefrequency and rotated as per thgenerated by uinstantaneous sc

The point wherevector, is compuand mathematicrealized by usiThese methods dand also dependterrain. These m

The size of rastraster cell is smfeasible to storeand algorithms aThe tiles whichknow by the algthe swath of a pa

3.6 Error intro

LiDAR data sudifferent kinds and orientation oby sensor propprovide facility version of simintroduced in thein X, Y and Z derror in X direct

i iT tX X N= +

Where XTi is the

with different vcoordinates. It follow normal errors it is ensucorrelation of eknown from fielsensors. The sim

3.7 Output

This module haoutput files asdescription abou

X, Y, Z: This module wrThis can be dispVoxel, Terascan

LAS 1.0 foLAS 1.0 is ASPintention of the

flight lines depmeters (i.e. swatfinal points of etotal numbers oAR data generati

ponent generatesending on the velocity of th

he direction of using trajectory can angle and att

e laser hits the teuted by solvingcal surface or ing specially fdiffer for mathemd upon the bas

methods are incor

ter data becomemall. Therefore,e entire data in are designed to lh fall under the gorithm. It reads articular flight li

oduction in dat

uffer from syst(Huising and Peof platform and pagate in final

for introductiomulator. In pres

e terrain coordindirections separation is shown bel

2( , )X XN μ σ

e X coordinate vvalues of paramis assumed that distribution. F

ured in algorithmerror. The parald experience anmulator facilitate

as several sub m in case of a

ut the modules is

ites generated Xplayed in many n etc.

ormat: PRS LIDAR data

data format is t

pends on the dirth width and pereach flight are cof flight lines coion.

s trajectory coordistance of fli

e aeroplane. Tflight lines. Thcoordinate and

titude.

errain, followingg the intersection

rasterized terraformulated nummatical surface aic equation emprporated in separ

s very large, in , to solve intersmemory. Speci

logically tile the swath of the flonly those tiles

ine.

ta

tematic and raereira, 1998). Ein angle and rancoordinates. It

on of these erroent version a

nates computed iately. The systemlow:

alue with error. meters are usederrors in X, Y

Further, when inm that there is ameters of this

nd are reported bes variation of th

modules which gan actual LiDAs given as follow

X, Y, Z data in th3D display soft

a exchange formo provide an op

rection of flightrcentage overlapcalculated by theover whole area

rdinates for eachight line, firing

The trajectory ihe laser vector id corresponding

g the above lasen of laser vectoain. Solution i

merical methodsand raster terrainployed to createrate modules.

particular whensection it is noial data structuredata.

light line can bewhich fall unde

andom errors oErrors in positionnge measurement

is proposed toors in the futurenormal error iin the above step

m for introducing

(2)

Similar systemd for Y and Zand Z directionntroducing theseno spatial autodistribution are

by the vendors ohese parameters.

generate differenAR survey. Thews.

he ASCII formattware e.g Surfer

mat standard. Thepen format which

t, ). e a

h g s s g

r r s s. n e

n ot e

e r

of n t o e s

p g

s Z s e -e f

nt e

t. r,

e h

allows difwhich a vmodule wra standard LAS

This formawrites gen Time

This writeASCII form Time

This writecolumn wi Time

This writeASCII form Repo

This writeinformatiovelocity, atime, firinetc.

F

Simulated 14 secondcompare sthese two exhibit the A hypothesurface andrag-drop created. ThA view ofsurface fea

fferent LIDAR vvariety of LIDrites data in this format.

1.1 format at is the next ver

nerated data in th

, Attitude es time, Roll, Pmat.

, Acceleration es time and accise in ASCII form

, Error s time and errormat.

rt es a report file inon regarding Lialtitude, number ng frequency, sc

igure 9. LiDAR

4. RESUL

attitude and accds (Figure 10). simulated data wrepresent two di

e random nature

etical terrain (40nd is populated

facility of softhe flat surface a

f this is shown inature of surfer.

vendors to outpAR software vformat so that t

rsion of LAS 1.his format.

Pitch and yaw

eleration valuesmat.

r introduced in X

n ASCII formatiDAR data genof flight lines, s

can frequency, s

data output men

LT AND DISCU

celeration are shThough, it is s

with any set of ifferent populatiof parameters as

00 m by 600 m)with building l

tware different and objects on ton Figure 11, wh

put data into a vendors can usethe generated dat

0 format. This m

data column w

s in X, Y, Z dir

X, Y, Z column w

t which containsneration processensor type, totalswath and terrai

nu and options

USSION

hown for a durastatistically diffif actual flight dions, the formers in any normal

) is created overlike shapes. Bytype of buildinop of it are rast

hich is displayed

format e. The ta is in

module

wise in

rection

wise in

s detail s. e.g. l flight in type

tion of cult to ata, as amply flight.

r a flat y using ngs are erized.

d using

LFK3TpFa

Figure 10. A

Figure

Lidar data are Flight velocity: KHz; Scan frequ3; Overlap 10Terrascan softwpresented for thFigure 13. To saltitude value is

ttitude (top) and

11. Surfer surfac

generated for th60 m/s;Altitud

uency: 48Hz; Sc%. Resulting L

ware and displayhe sake of spacshow the effecttaken.

d acceleration (bo

ce of the created

his terrain with de: 190m; Firingcan angle 500; NLiDAR data ayed. Only few ce as shown int of multiple fli

ottom) values

d terrain.

the parametersg frequency: 20o. of flight lines

are imported inviews are being

n Figure 12 andight lines, smal

s: 0 s: n g d ll

Profile viewhich is hwhere datshadow is not capturmarked bycan be undblack areaserves as height, wh

Figure 12.

Figure 13.

Figure 14 n

ew of the generahaving three flighta are capturedalso evident. N

red) are shadowy white oval is nderstood in profa is not being coan example of p

hich can be unde

. Lidar data displ

Lidar data displThe profile is sh swaths. The bu

. Effect of high no attitude varia

ted LiDAR dataht lines. This sh

d, while the intNot all black arews. For examplenot fully capturefile (also markedovered by eitherpoor choice of rstood by simula

lay in Terrascan

lay in Terrascanhown along withuilding within ov

(a)

(b) attitude on comp

ation. (b) with hig

a is shown in Fighows various buiterplay of objecas (i.e. where da

e, the roof of bued. The reason fd by white ovalr flight line. Thscan angle and ator.

n for the above su

- profile view. h the flight lines val is not measur

plex building. (agh attitude varia

gure 13 ildings ct and ata are uilding for this .). The

his also flying

urface.

and red.

a) with ation.

T

aasrdlfoTiF

Tissaatuicpfpdeup

The zoomed out14 (a) and (b)) oattitude variationattitude effects ospread of point crespect to flight data points are aline, while no daflight lines are soverlap of two fThe generated rain surfer. The LiFigure 16)

Figur

Figure 16. Lid

The present versinterface so thasimulates the psensor as well according to hiactual LIDAR. Dterrain by settinguseful to generais also useful icapture process parameters and flight planningparameters, fligdeveloped usinextended furtheused to developprogramming a

t view of a compof the same placens. These figureon the point cloucloud depends online and parame

available on the vata points are caphow by arrow. T

flight lines. aster for a fractaiDAR data is gen

re 15. Fractal sur

dar data of fracta

5. CONC

sion of software at the user can process of LiDAas the user hass choice whichDifferent data seg the different p

ate LiDAR data fn a classroom and understanditheir errors. Fur. Before actua

ght direction etcng object-oriente

r very easily. JAp the software s well as havin

plex building (she with no attituds give understan

ud. This also shon the location of

eters chosen for svertical wall facptured on the othThe figure also s

al surface is shownerated for this f

rface display in s

al surface display

CLUSION

offers a user frieuse this softwa

AR data collects freedom to seth is not possibleets can be generaparameters. The for research to tefor demonstrati

ing the effect of rther, it can be ual flight the ec can be seen. ed methodologyAVA programmas it supports

ng excellent GU

hown in Figure e and with high

nding that how ows that the f the object with sensor. LiDAR ing the flight her wall. The shows the

wn in Figure 15 fractal (shown in

surfer.

y in Terrascan

endly GUI basedare with ease. Ition for LIDARt the parametere in the case oated for the samesimulator can beest algorithms. Iing LiDAR dataflight and sensouseful in LiDAReffect of sensoThe software iy so it can be

ming language iobject oriented

UI features. The

n

d It R s f e e It a r

R r s e s d e

platform inmultiple op The preseincorporatIMU dataconcepts oincorporatmultiple re ReferenceBeinat, Astochastic images, Insensing an Marciniakencyclopedfor Softwa Holmgreneffect of liand canop29(5), pp. Husing, Eestimates systems fPhotogram Lohani, BlaboratoryPhotogramLaser ScaSeptember Kukko, Asystem anameasuremand SiviLa Lohani, B.AltimetricArchives ofInformatio Lohani, B.data collecconference Queija, V.U.S. geolopp. 5-9. Sun, G. anforest canosensing, 38

5.1 Ackn

This work

ndependency ofperating systems

ent version ofte more faithfulna collection anof separation of ting atmospherieturns, etc.

es A., and Crosilla

model for opnternational Arcnd spatial inform

k, J.J., 2001, Prodia of software are Research, Un

, J., Nilsson, M.idar scanning anpy closure. Can623-632.

E. J. and Pereiraof laser data

for topographicmmetry & Remot

B., Mishra, R. Ky: LiDAR simmmetry , and Ranning 2007 anr 12-14.

A. and Hyypa Jalysis and algor

ments, ISPRS Waser 2007, Espoo

., Reddy, P., and LiDAR Simula

of the Photogramon Science, XXX

., 2001, Airbornction: Issues ande MAPINDIA-20

R., Stoker, J. Mogical survey app

nd Ranson, K. J.,opies, IEEE tran8(6), pp. 2617-2

nowledgements

is supported un

f JAVA made ths.

f simulator is ness by introducnd their integra

GPS antenna, laic effects and

a, F., 2002. A ptimal registratichives of photog

mation sciences,

ocess models in engineering. W

niversity of Calif

, and Olsson, H.ngle for estimationadian Journal

a, L. M., 1998,acquired by vac applications, te Sensing, 53(5)

K., 2007, Genermulator, InternRemote Sensin

nd SilviLaser 2

., 2007, Laser rithm developme

Workshop on LAo, September 12

d Mishra, R.K., 2tor: An educatio

mmetry, Remote SXVI(6), Tokyo, J

e altimetric LiDd application., Pr001, 7-9 Februar

M., and Kosovichplications of LiD

, 2000. Modelingns. On geoscienc2626.

der RESPOND p

his software to r

being modificing separate GPation; by introdaser head and IN

d by facilitatin

generalized faon of LIDAR grammetry an r34(3/B), pp. 36-

software engineWalt Scacchi, Infornia, Irvine.

., 2003. Simulation of mean tree l of Remote Se

, Errors and acarious laser sca

ISPRS Journ), pp. 245-261.

rating LiDAR dnational Archivng XXXVI(3)W2007, Espoo, Fi

scanner simulatent: A case withASER Scanning-14, 2007, Finla

2006, Airborne on tool, InternatiSensing and SpaJapan.

AR for topograproc. of Internatiory 2001, New De

h, J. J., 2005, RecDAR, PE&RS, 7

g Lidar returns fces and remote

programme of IS

run on

ied to PS and ducing NS; by ng for

actored range

remote -39.

eering, nstitute

ing the height

ensing,

curacy anning nal of

data in ve of

W52 of inland,

tor for h forest g 2007 and.

ional atial

phic onal elhi.

cent 1(1),

from

SRO.