M. Batista, T. Magister, L.

Bogdanovic: Computer Based Road Accident Reconstruction Experiences

MILAN BATISTA, D. Se. Traffic Safety

E-mail: [email protected] Preliminary Communication
TONE MAGISTER, M. Se. U. D. C.: 656.087:351.81
E-mail: tone. [email protected] Accepted: Oct. 18, 2004
LEON BOGDANOVIC, A. Eng. Approved: Mar. 1, 2005
E-mail: leon. [email protected]
Univerza v Ljubljani
Fakulteta za pomorstvo in promet
Pot pomorscakov 4, SI- 6320 Portoroz, Republika Slovenija

COMPUTER BASED ROAD ACCIDENT

RECONSTRUCTION EXPERIENCES

ABSTRACT The foremost computer software intended for sim-

ulations of automobile collisions (SMAC- Simulation
Since road accident analyses and reconstructions are in- Model of Automobile Collision) was developed at the
creasingly based on specific computer software for simulation aeronautical laboratory Calspan at the Cornell Uni-
of vehicle d1iving dynamics and collision dynamics, and for versity in 1974. As a preprocessor of the SMAC pro-
simulation of a set of trial runs from which the model that best gram another program (CRASH- Calspan Recon-
describes a real event can be selected, the paper presents an struction ofAccidents Speeds on Highway) was also de-
overview of some computer software and methods available to
signed in the same laboratory and was later developed
accident reconstruction experts. Besides being time-saving,
when properly used such computer software can provide more
as a standalone program CRASH3 [4). While SMAC
authentic and more trustworthy accident reconstruction, there- is intended for collision simulations, which include ve-
fore practical experiences while using computer software tools hicle motion before, during and after the collision,
for road accident reconstruction obtained in the Transport CRASH calculates the speed change (delta v) of vehi-
Safety Laboratory at the Faculty for Maritime Studies and cles during the collision and for their estimated mo-
Transport of the University of Ljubljana are presented and dis- tion after the collision from known damage and posi-
cussed. This paper addresses also software technology for ex- tions of the vehicles before and after the collision.
tracting maximum information from the accident photo-docu- With the appearance of personal computers, the use
mentation to support accident reconstruction based on the sim- of simulation programs became everyday practice for
ulation software, as well as the field work of reconstruction ex- road-accident reconstruction experts. In 1986 com-
perts or police on the road accident scene defined by this tech-
mercial versions EDSMAC and EDCRASH ap-
nology.
peared [14), and in 1994 the software named m-smac
[15] and later on in 1996 the programs WinSMAC and
KEYWORDS
WinCRASH became available [13). In the mid-nine-
ties the programs such as CARAT- Computer Aided
road accident, investigation, reconstruction, simulation, safety
Reconstruction of Accidents in Traffic and PC-Crash
[9) were developed in Europe. Unlike SMAC and
CRASH, these two programs are based on the model
1. INTRODUCTION of impulsive collision.
The reliability, accuracy and trustworthiness of an
The road accident reconstruction has been defined accident reconstruction depend on the quality and
as the effort to determine from whatever information quantity of the data available from which to make an
available how the accident happened at one or several evaluation. Reliable input data has to be provided for
points in time [12, 6). As scientific work under estab- successful analysis performed by computer software
lished practical conditions, an accident reconstruction tools for collision simulations as well.
is based upon a set of analyses considering: collision, By the time an accident reconstruction expert is
injury, accident and injury avoidance, and accident sought the only available data consists ofthe police re-
causation. To perform a majority of analyses needed port, the police sketch of the accident scenario, min-
for the accident reconstruction computer software utes ofthe statements in court, and photographs of the
tools are nowadays available to the reconstruction ex- scene with record of the skid marks, vehicle damage,
perts. and involved vehicles still on their positions at rest.

Promet-Traffic-Traffico, Vol. 17,2005, No. 2, 65-75 65

 M. Batista, T. Magister, L. Bogdanovic: Computer Based Road Accident Reconstruction Experiences

Namely, by the time a reconstruction expert investi- vehicle is a rigid body on elastic suspensions or simply
gates the scene of an accident, the vehicles had been suspensions are ignored. The motion of vehicle is de-
moved from the scene, usually already repaired, the fined by the forces acting on the vehicle tires. For the
skid marks on the scene of an accident have faded and sake of definiteness, we provide the basic equations
the debris has been washed away. for two-dimensional vehicle motion simulation on a
The data required for accident reconstruction are horizontal ground with air resistance neglected [1] .
usually available only in the photographs taken by the The basic vehicle geometry data and its position on
police on the day of the accident and can be extracted the plane are shown in Figure 1. When the position of
from them. The first computer programs intended for the center of gravity of the vehicle is not known the
photogrammetric analyses of accident scenes were de- 50:50 distribution between front and rear axes are as-
veloped in the mid-eighties [2]. Nowadays, excellent sumed.
special software such as PC-Rect [7] and commercial The differential equations which describe the vehi-
of the shelf software is available for processing photos cle motion are:
of the accident scene. The use of such software en-
aDro .
ables reconstruction expert to create an accident dt= Vx COS1./J-Vy Slll1./J
scene diagram to scale which includes the location and
the length of skid marks, vehicle positions at rest, and dY
impact location from an accident photo-documenta-
dt0 = Vx sin 1./J- Vy COS1./J
tion. Furthermore, a digital model of a vehicle damage d1jJ
sustained in an accident can be extracted from a pho- -=w
dt (1)
tograph with the use of off-the-shelf software such as
PhotoModeler [8]. dvx
- 1
-=- (2: F ) +wv
dt m x Y
The imbued purpose of this article is to describe
1 (
some relevant computer software tools for accident si-
mulation and reconstruction, to provide pros and con-
dvy
-=-
dt m
2:FY ) -wv x
tras for their usage, and their limitations are also dis- dw =....!__~M
cussed. Everyday experiences of a reconstruction ex- dt J z LJ z
pert namely show that when one particular computer
software is relied upon the exclusion of others or when where m is the total vehicle mass including mass of
accident reconstructions relies solely upon simulation, the passengers, J z is the inertial yaw moment of the
erroneous conclusions can be drawn. It has to be stres- vehicle around its center of gravity, Vx and vy are com-
sed that the presented article is not an advertisement; ponents of the velocity of vehicle's center of gravity
the intention of its authors is to provide their practical expressed in its own coordinate system, w is the angu-
experience while using computer software tools for lar velocity of the vehicle, LE'x and LE'y are forces in
road accident reconstruction gained in the Transport the direction of vehicle coordinate axes and LMz is
Safety Laboratory at the Faculty for Maritime Studies sum of yawing torques around the center of gravity
and Transport of the University of Ljubljana. of the vehicle. The system of equations (1) is inte-
grated with the Runge-Kutta method with fixed incre-
ment or even with low order Euler method. The vehi-
2. PRIMARY RECONSTRUCTION cle inertial moment can be calculated simply by the
TOOLS- TRAJECTORY AND assumption that the vehicle is a homogenous plate
2 2
COLLISION SIMULATIONS J z = (/ + w )m I 12 [7] or using the empirical for-
mula J z = 0.1269xm x wxl [9].
The calculation of vehicle motions in the accident In order to calculate the wheel forces, the vertical
reconstruction software is based on assumptions that a load on a particular wheel has to be calculated first .
y

~
w

a ,, ~
y

w t, I 4

3
r,
~
2

1
It
I X

Figure 1 -The variables of vehicle dimensions and position

66 Pro met- Traffic- Traffico, Vol. 17, 2005, No. 2, 65-75

 M. Batista, T. Magister, L. Bogdanovic: Computer Based Road Accident Reconstruction Experiences

For the model of a vehicle with suspensions, the verti- collision simulation is basically distinguished between
cal forces on a particular wheel are: those based upon the impulsive model, and the others
based upon the model of vehicle deformations.
1( b h ) hKt
Fzl =2 a+bmg- a+bLFx + t K +t K LFy
f f r r
The software algorithm for the reconstruction of a
vehicle collision is based on the impulsive model
1( b h ) hKt which comprises the laws of conservation of linear and
Fzz = 2 a+b mg- a+b LFx - t K +t K LFy
f f r r
angular momentum. The fundamental supposition of
the impulsive model is that the collision is instanta-
F
Z3
- _!_(_b_m - _h_ "F
- 2 a+ b g a+ b L.. x
)+ ft K
hKr
f +t r K r
"F
L.. Y
neous (i. e. the vehicles do not change their positions
during the collision phase). The impulse model is
mathematically described as follows [1 ], [9]. First the
F 4 - _!_ (-b-
m - ___!!__ " F ) -
z - 2 a+ b g a+ b L.. x
hKr
t K +t K
f f r r
" F
L.. Y
local coordinate system is centred in the point of as-
sumed vehicle impact, which is named the impact
(2) point P. The coordinate system has axes n, t directed
where g is Earth gravity acceleration, h is vertical posi- perpendicular and tangential to the assumed contact
tion of the vehicle centre of gravity and KJ, Kr are the plane (Figure 2).
front and rear suspension stiffness respectively. When
the suspension of the vehicle is ignored, the vertical
loads on wheels are simply static loads applied on
them:
b
F -F - mg
z,l- z,2- 2(a+b)
(3)
b
F z,3-
-Fz,4-
- (a+b) mg
2
The components of the particular force on the k-th
wheel are calculated from the following formulas: Figure 2 - Oblique collision used in the impulse model
Fx,k = Fwx,k cos o k - Fyw,k sin o k
(4) The conservation of linear and angular momentum
Fx,k = Fwx,k sinok +Fyw,k cosok yields:
where ok is the wheel steering angle and Fwx, k and Fwy, + - In
vln -vln = -
k (k = 1,... , 4) are the tyre longitudinal and lateral ml
forces expressed in the coordinate system of a wheel.
+ -
The total force applied on a wheel cannot exceed the v11 - vlt = -It first vehicle
available contact friction between the tyre and the sur- ml
face of the road: l1z<wt- w1) = InYl- Itxl
~F_!,k + Fy~,k $. flk Fz,k (5) (8)
+ - In
where flk is the friction coefficient between road sur- Vzn -vzn = - -
face and the k-th tyre. Longitudinal force Fwx, k on the mz
k-th wheel is defined by driving simulation (i.e. by de- + - It
v2 - vz1 = - - second vehicle
fining driver acceleration or breaking and road condi- t mz
tion). It is given by:
lzz(wt -w1)= -InY2 +Itxz
Fxw,k =8kflkFz,k>-1$.8k $.1 (6)
On the other hand the lateral force on the k-th
wheel depends on its sideslip angle ak which is defined where superscript + is used to denote unknown values
as the difference between the direction of velocity vec- after collision and -for values before collision, I 11 and
tor of the wheel centre and its steering angle and for its I 1 are components of impulse in the normal and tan-
calculation various models are used. The simplest is gent direction to the contact area whilex 1,yl, andxz,yz
the linear model: are coordinates of the centre of gravity of the first and
second vehicle respectively in the local coordinate sys-
(7) tem. To retrieve a solution ofthe system (5) two addi-
tional assumptions are needed for the unknown veloc-
While the vehicle trajectory calculation before and ities after collision and two components of the im-
after the collision in an accident reconstruction soft- pulse. Therefore, two cases are distinguished. Where
ware is exclusively based on rigid body dynamics the there is no relative motion between vehicles at the

Promet- Traffic- Traffico, Vol. 17, 2005, No. 2, 65-75 67

 M. Batista, T. Magister, L. Bogdanovic: Computer Based Road Accident Reconstruction Experiences

point of full impact, the velocities at impact point dur- In smac type programs [7] the crush pressures are
ing compression phase of impact are the same for both computed rather before than the crush forces. To do
vehicles. For the case of the sliding impact the as- that a set of equally spaced radial vectors from vehicle
sumption is that both vehicles reach the same veloci- centre of gravity are established first (Figure 4). Each
ties at impact point in normal direction, while tangen- vector represents a potential crush pressure against
tial direction of impulse is calculated from the Cou- the vehicle. To calculate the crush pressures the inter-
lomb law / 1 = pJn where,u is the coefficient offriction section between the vehicles has to be confirmed first,
between vehicles. and then all the potential radials are processed. The
The basic assumption that the vehicles do not first intersection of particular radials is confirmed and
change their positions during the collision phase rep- then two additional vectors are established. One ra-
resents the major disadvantage of the impulsive dial is drawn from the centre of gravity of the base ve-
model. Considering vehicle collisions such a supposi- hicle to the contact point and the other radial from the
tion should be treated only as rough approximation, centre of gravity of the other vehicle to the contact
because of the fact that vehicles can substantially point. The pressure PiU) at i-th point onj-th vehicle is:
change their positions during the collision phase. Fur- p~j)=k~j)(p~j)
t t t,max
-p~j))
1
(10)
thermore, before using such a model, a set of parame-
ters should be known to the reconstruction expert where Pi, maxUJ is the vehicle perimeter and p/j) is the
such as the vehicle position during the impact, the im- contact point radius. The program then iterates the
pact point and the direction at which vehicles collided. p/i) until p/f) and p/2) are balanced to the prescribed
Since in practice this data are often not available, tolerance. When the i-th radial Pi is in the restitution
some presumptions have to be adopted (e. g. impact phase it is enlarged by:
duration, speed loss at impact - EES (Energy Equiva- P p,t = c Pmax,i +(1- c)pi (11)
lent Speed)), otherwise the determination ofthese pa- where Pmax, i is the maximum i-th radial length and
rameters has to be left to the more or less educated c = 1- of I om simulate restitution instead of the co-
judgment of the reconstruction expert. If caution is efficient of restitution e.
not used, all these facts can ultimately lead to a recon-
struction which is not objective.
The disadvantages of the impulsive model dis-
cussed can be overcome with the use of the model
based on vehicle deformations sustained during the
collision. Contact forces between the vehicles are also
considered in the dynamic equations of such a model.
When impact occurs, the model calculates the contact
forces on the basis of vehicle mechanical properties
and consequently considers them in its equations of
motion. In this way the vehicle motion in the collision
phase is entirely described without intervention of the
reconstruction expert, which results in a more objec-
tive reconstruction of the accident.
Figure 4 - Computing the crush pressure
Usually, in the deformation based models the lin-
ear relationship between normal collision force F and When pressures and deformations are calculated,
the deformation is assumed with K as the coefficient of the normal force to a chord segment joining two con-
o
stiffness of the vehicle and as the deformation or the secutive deflected radial points is calculated (Figure
crush depth of the vehicle (Figure 3): 5). The normal force in i-th segment is:
F= Ko (9) 1
F,1,i = 2_(Pi + Pi+I)t:lii (12)
F where Pi is pressure at the beginning and Pi+l at the
end segment point, and /1/i is segment length. The tan-
gential friction force in i-th segment is:
Ft,i = ,U v Fn,i (13)
where .Uv is the inter-vehicle friction coefficient. In or-
der to compute the velocity difference, note that the
tangential velocity at the middle of i-th crushed seg-
Figure 3 - The force-deflection relationship where 1 a ment for the base vehicle is (Figure 5):
am
is the final (residual) deflection, is the maximum
v1,i = (v x- w1Ji) cos <Pi+( vy + w~) sin <Pi (14)
deflection

68 Promet- Traffic- Traffico, Vol. 17, 2005, No. 2, 65-75

 M. Batista, T. Magister, L. Bogdanovic: Computer Based Road Accident Reconstruction Experiences

y
~

y 1: v,,;

X X

Figure 5 - Forces and velocity at crushed segment

collision and with the estimated location of a point at

which collision occurred. The reconstruction expert's
task is then to search for the calculated solutions for
the set of positions describing vehicle motion and de-
formations of the vehicles which best suit the real
data. This is done via changes and iterations of the ini-
tial vehicle positions and their properties as well. It
has to be stressed that there is a large amount of inputs
to be made; 27 parameters defining both vehicles in-
volved in an accident (dimensions, mass, yaw moment
of inertia, chassis stiffness, tyre stiffness) and 6 param-
eters defining initial conditions (position, angle of ro-
tation, initial speed, initial angular speed) have to be
defined. If also the friction coefficient and the tabu-
Figure 6 -The results of m-smac computer simulation lated data describing the forces on tyres and the ac-
of a two-vehicle collision; individual positions of the tions taken with the steering wheel are taken into ac-
vehicles are plotted in 0.25 s intervals.
count, then the search for a suitable solution with
m-smac can be very exhaustive. This time-consuming
where ~i = O.S(gi+gi+l) and 'ifi = 0.5(1Ji+1Ji+l) and
procedure can be abbreviated by some manual calcu-
if>i is the i-th segment direction. A similar formula is
lations before the use of the program, which serve to
used for the other colliding vehicle. Once both veloci-
estimate the initial values (for example WinCRASH).
ties at contact point are known the velocity difference
Another program based on impulsive model in
can be calculated.
widespread use is PC-Crash. It enables collision simu-
Once normal and tangential forces are known,
lations up to 32 vehicles. Beside 2D simulations, it also
they must be transformed from the crushed segment
provides 3D simulations of vehicle collision and roll-
coordinates to vehicle coordinates by:
over, simulations of vehicle collision into fixed barri-
Ft;,i = Ft,i cos if> i + Fn,i sin if> ers, simulations of vehicle-pedestrian and vehicle-mo-
(15)
F17,i = F1,i sin if>i- Fn,i cosq> torcycle collisions and also simulations of passenger
motion in the vehicle. Some examples of use in recon-
The yaw moment caused by collision forces is then:
struction of real accidents are shown in plates 1, 2, 3
2_ M= 2_ (~ J 17,k -1JFt;,k) (16) and 4.
An dample of a program based on the deforma- Unlike m-smac, PC-Crash has a simple and user-
tion model is m-smac [8]. This program allows 2D sim- -friendly graphical interface, which enables fast defini-
ulations of two vehicles collisions, collisions of vehi- tion of vehicles due to its vehicle database, input of
cles into a fixed barrier and collisions of vehicles into a different surfaces and settings of initial vehicle posi-
narrow barrier (pole, tree). The data is input via ta- tions in a simple way according to the police sketches
bles, while the results are given in tabular or graphical from the spot or rectified photographic material from
form. The main advantage of this program is its possi- the scene of the accident. Instead of time-consuming
bility to allow comparison between the calculated and tabulated inputs for the driver's actions such as brak-
actual deformation of the vehicles at collision. Figure ing, acceleration, steering, as in the case of m-smac,
1 shows the use of the program (analysis of a two-vehi- the reconstruction expert can benefit from graphical
cle collision). inputs (e. g. vehicle course) and also from automated
In the m-smac program it is common to start with search of best initial conditions. It can be said that with
the known positions of the vehicles at rest after the PC-Crash the reconstruction expert almost drives the

Promet- Traffic- Traffico, Vol. 17, 2005, No. 2, 65-75 69

 M. Batista, T. Magister, L. Bogdanovic: Computer Based Road Accident Reconstruction Experiences

Plate 1 - PC-Crash simulation of a collision of a vehicle Plate 2 - The PC-Crash simulation of an occupant's
into side-road barrier-type stone and the initial motion motion before impact after skidding with locked
of a vehicle over side-scarp at the speed of 37 km/h. wheels; interval between frames is 0.5 s.

car involved in the accident when simulating the acci-

o.o. 0.051 0.101 dent dynamics. The main disadvantage of the program
is slow calculation in the case of 3D models and huge
amount of settable parameters.
As another accident reconstruction and simulation
tool the WorkingModel [11] can be used, but for the
analyses of certain types of accidents. The software is,
namely, intended for simulations of the motion of any
mechanical system in a 2D plane. Therefore, as the ac-
cident simulation platform it requires that everybody
and everything involved in an accident and an accident
Plate 3 - PC-Crash simulation of a pedestrian's event i. e. vehicles, humans and other objects involved
rebound from a car hood shown from in a road accident as well as their motions are 'trans-
the driver's perspective lated' into mechanical systems by the reconstruction
expert him/herself. Plate 5 shows an example of its use
in an analysis of an accident.
2 3 4 5

6 7 8 9 10

Plate 4 - PC-Crash simulation of a vehicle-cyclist collision at vehicle speed of 40 km/h,

and cyclist speed of 10 km/h.

70 Promet- Traffic- Traffico, Vol. 17,2005, No. 2, 65-75

 M. Batista, T. Magister, L. Bogdanovic: Computer Based Road Accident Reconstruction Experiences

100 600 1100 1600

~ ~ ~ ~
200 700 1200 1700

~ (; ~ ~ ~
1800
80~
300
13~
~ G~
1900
90~
400 1400

~ ~ ~
1500 2000
10~
500

~ ~ ~
Plate 5 - Working-Model simulation of a vehicle collision into a motorcycle, the simulation of the motorcycle
driver and the passenger motion after the initial impact of a vehicle (vehicle speed is 72 km/h,
while the motorcycle speed is 36 km/h; intervals between frames are in .us)

3. SECONDARY RECONSTRUCTION point P(xp, y p, z p ), position C(xc, Ye. zc ), orienta-

TOOL tion (rotation around x-axis 1./J, around y-axis </J, and
around z-axis ~) and the focal length f of the camera
should be given as well as the photograph height h and
A prerequisite for the accident reconstruction ex- width w, and coordinates of the point in the photo-
pert is the creation of accurate accident scene diagram graph A(uA, vA) expressed in two-dimensional u-v
to scale from photographs taken when the accident coordinate system of the photograph. When point P
scene was investigated by the reconstruction expert. If lies on a flat surface on the x-y plane of the three-di-
photographs taken on the day of an accident are avail- mensional space (Figure 7) then its coordinates are
able, an accident scene diagram with details of the lo- P(x p, y p,O). The camera should be positioned on the
cation and the length of skid marks, vehicle positions z-axis above the origin of the three-dimensional coor-
at rest, and impact location can be created. Software dinate system and its coordinates are then P(O,O, z p)
supporting the process of inverse camera projection is as shown in Figure 7.
applicable when little or nothing is known about the If the visual reference point R lies in the middle of
camera that was used to take accident photo-docu- the photograph, and if the camera is not rotated
mentation. To transform a surface of an object in true around z-axis, therefore ~ = 0, and the axis .; through
three-dimensional perspective of the photograph into reference point R from camera C lies on the y-z plane,
the plan two-dimensional view drawing (Figure 7, and then according to Figure 7, its coordinates in the
Plate 6) the applicable software should support the three-dimensional x-y-z space are:
process of photographic rectification. When the soft- XR= 0
ware for photographic rectification is used in the pro-
cess of an accident reconstruction then a software al- YR =- f COS'I./J (17)
gorithm and its characteristics define the field work of z R = z c - f sin 1./J
the reconstruction expert investigating the accident Vectors u and v defining the two-dimensional
scene, and a policeman on the accident spot as well. plane u-v of the photograph must be orthogonal to
Rectification is a procedure or geometrical tool each other ul_v, they also must be orthogonal uj_£',
which determines the relationship between the three- and v l_£ to the axis .; through reference point R from
-dimensional space and the two-dimensional image in camera C (Figure 7). When the described conditions
the photograph with objects in true perspective. To for the vectors u and v are met and when camera is not
convert a point in the two-dimensional photograph rotated around y-axis, therefore <P = 0, and around
A(uA, v A) into the corresponding three-dimensional z-axis, therefore ~ = 0, then vector u is orthogonal to

Promet- Traffic- Traffico, Vol. 17, 2005, No. 2, 65-75 71

 M . Batista, T. Magister, L. Bogdanovic: Computer Based Road Accident Reconstruction Experiences

C(O, 0, zc)

A* : 0 y

V
A(uA, vA) X

P(xp, yp,O}----__ A(uA , v.J

i
or A(xA, YA• z.J
u ~ ~ ~--- 0 y
w R(xR, YR.zR) C(O, O, z
A(uA, v.J ... the point on the photograph
P(xp. YP• 0) ... the point on the surface
C(O, 0, zc) ... the position of the camera photograph
R(xR, YR· zR) ... the view reference point

Figure 7 - Transformation of a photograph point A into a surface point P

they-axis u .ly and also to the z-axis u .lz, therefore, as graphic projection is a remapping of the image on a
shown in Fig. 7 its projections in the three-dimen- flat surface. Adequate software tools for photographic
sional x-y-z space are ux = 1 and uy = uz = 0; since processing, such as PC-Rect, which is an additional
vector v is orthogonal to the x-axis v .lx its projections module to the program tool PC-Crash for the recon-
in the three-dimensional x-y-z space are v x = 0, struction of vehicle motion dynamics, offer practically
vy = - sin VJ, and vz = cos VJ (Figure 7). automated inverse camera projection and rectifica-
Based on (17) and presumptions described the tion and are therefore indispensable in the analyses
point A( uA, v A ) from the two-dimensional u-v pho- and reconstruction of traffic accidents [9]. The work
tograph can be defined in the three-dimensionalx-y-z with the rectification software as PC-Rect is not de-
space as A( x A, yA, z A ) with coordinates: manding but might become time-consuming. If the
XA = WLlB photographs of the surface (meant for rectification)
YA =-fcosVJ-hvssinVJ (18) are taken with an analogue camera, their digitaliza-
tion is needed. When scanning, the coordinate sys-
ZA = Z C - f sin 1/J + h VB COS 1/J tems of the scanned picture and the original photo
When the point in the photograph is defined in the have to be in accordance.
three-dimensional space A( x A, yA, z A ) then the co- The rectification procedure starts already at the
ordinates of the corresponding three dimensional sur- scene of the accident, the surface intended for rectifi-
face point P(x p, yp ,O) can be calculated based on the cation has to be prepared scenically even before the
similarity between triangles !::,COP and /::,AA *p shown photographs are taken. This preparation is dependent
in Figure 7 and considering (17) and (18) it follows: on the working principle or geometrical model of
Xp=W2U2 1 transformations from two-dimensional plane into
B fsinVJ-h vs cosVJ three-dimensional space of the software tool which is
f COS 1/J + h VB COS 1/J used for rectification.
y =-wu (19) In order to rectify an image of a surface in the pho-
p B f sin 1/J + h VB COS 1/J
tograph and in order to determine the camera charac-
Zp = 0
teristics at least four points have to be defined on that
When relationship between the point of an object surface (of the accident scene for example) first. The
in the three-dimensional space and the point of the reference points can be defined either with drawing on
same object in the two-dimensional image in the pho- the surface or with displacing objects of the known
tograph (19) is determined an orthographic projection height. Those reference points should form a polygon
of the surface can be generated; where an ortho- with sides of at least 15 m. Field work showed that ob-

72 Promet- Traffic- Traffico, Vol. 17, 2005, No. 2, 65-75

 M. Batista, T. Magister, L. Bogdanovic: Computer Based Road Accident Reconstruction Experiences

long polygons disfigure rectification and produce non- those lengths differ too much, the composition of the
-negligible rhomboid shape deviations of the rectified two photos is not possible due to their disfigurement.
photograph from real geometry with lateral errors in- An arbitrary number of reference points has to be
creasing in the direction of the camera objective. Lon- defined on a surface which will be later made up of
ger sides of a polygon result in difficulties while taking multiple planes rectified separately. After the mea-
photos on the spot. The surface of interest in the pho- surement of all the combinations of reference dis-
tograph for rectification must also be identified in the tances among the reference points, the surface has to
process by choosing control points which will enclose be photographed in such a way that a single photo
this surface of interest into an arbitrary polygon. shows limited number of reference points; in case of
Within its boundaries an arbitrary polygon should PC-Rect one photograph should show only six points
contain only limited number of reference points; in at the most. Whenever possible the reference dis-
case of PC-Rect one photograph should show only six tance, which will be later in the process used for stitch-
points at the most. The software, namely, enables the ing of multiple planes rectified separately should ap-
input and processing of a maximum number of ten ref- pear on the photograph taken closest to the camera to
erence distances for the single rectification. Further- minimize deviations on the rectified flat surface. The
more, this arbitrary polygon should follow the edges of latter means that the area of concern should be photo-
the surface of interest as accurately as possible. The graphed from opposite directions.
deviations namely produce disfigurements of the rec- Beside the determination of the reference points
tified picture. and measurement of the reference distances on the
In order to make a plan view drawing of the surface surface intended for rectification, at least the focal
of a curved road, such a surface should be broken length of the camera, and height of the camera have to
down into as many pieces as reasonable to best repre- be determined for every photo according to (19). In
sent the curvature of the road because a rectification order to minimize deviations of the rectified picture
plane cannot be curved. Each piece can then be recti- from actual geometry, it is also recommended for ev-
fied separately and composed after rectification as ery photo that the inclination angle ljJ of the camera is
shown in Plate 6. The multiple planes should be iden- determined (Figure 7, and (19)). An automated in-
tified for separate rectification for any extensive area verse camera projection routine optimizes the posi-
or an area which is not planar and is intended for recti- tion of the camera objective with regard to the mini-
fication. Multiple planes rectified separately can be mum of four reference points defined within the arbi-
stitched together to form an extensive area or a sur- trary polygon of chosen surface of interest in the pho-
face of a curved road by means of adjacent reference tograph for rectification and calculates the missing pa-
distances of a pair of polygons from the adjoining rameters of the camera position when the two-dimen-
planes. A deviation condition of their rectified lengths sional plane of a photograph is misaligned with the
for a pair of adjacent reference distances exists. If three-dimensional coordinate system and missing ori-

Plate 6- Original photos of an accident scene and rectified picture of a blind curve (right);
the rectified picture is composed from three independently rectified oriQinal photos

Prom et- Traffic- Traffico, Vol. 17, 2005, No. 2, 65-75 73

 M. Batista, T. Magister, L. Bogdanovic: Computer Based Road Accident Reconstruction Experiences

entation parameters of camera rotation around y-axis There is no need to survey the scene with traditional
rp, and around z-axis ~ in such a way that the weighted survey instruments; therefore, there is no need to get
error of the reference distances is minimal. If the work into traffic, which can be dangerous and time-consum-
is sufficiently accurate, the weighted error of the PC- ing. One should always be aware that the process of
-Rect will be in the range between 0.3% and 3% [10]. rectification and its results are as good as the quality of
Reported error when photographs are rectified using the photographs and camera used.
PhotoModeler are in the range between 0.4% and
0.9% [2]. MILAN BATISTA, D. Se.
E-mail: [email protected]
TONE MAGISTER, M. Se.
4. CONCLUSION E-mail: tone. [email protected]
LEON BOGDANOVIC, A. Eng.
E-mail: lean. [email protected]
The comprehensive illustrations and descriptions
Univerza v Ljubljani
present the applicability and practicability of available Fakulteta za pomorstvo in promet
software for road accident reconstruction and reliabil- Pot pomorscakov 4, SI- 6320 Portoroz, Republika Slovenija
ity of its simulations. In the Transport Safety Labora-
tory at the Faculty of Maritime Studies and Transport POVZETEK
of the University of Ljubljana we do not prefer partic-
ular software package over another; they are used in- IZKUSNJE PRI RACUNALNISNO PODPRTEM RE-
dividually or in combination according to the type of KONSTRUIRANJU CESTNO-PROMETNIH NESREC
the accident reconstructed. It is advantageous and
sensible to substantiate any conclusions drawn from as Ker analize in rekonstrnkcije cestno-prometnih nesrec vse-
many different directions as possible. This cross- pogosteje podpira specializirana programska oprema, ki omo-
goca simulacijo dinamike voinje vozil in trkov, omogocajo pa
-checking renders additional weight to the final op-
tudi simulacijo niza poiskusov med katerimi je moc izbrati tisti
tion. model dogajanja, ki najverodostojneje opisuje resnicno doga-
Modern software tools become indispensable for janje V casu nesrece, podaja pricujoCi prispevek pregled neka-
the reconstruction of some types of traffic accidents. terih racunalniSkih programov in metod, ki so na razpolago
Instead of speculations and rough estimates, they of- izvedencem pri raziskavi cestno-prometnih nesree. Poleg tega,
fer more accurate calculations of the motion of the da je z njihovo uporabo moc prihraniti cas, omogoca lovrstna
participants in an accident. With the aid of these tools, programska oprema realnejse in verodostojnese rekonstntkcije
injuries to humans and damage to vehicles can be cal- cesto-prometnih nesreC, zato SO V clanku predstavljene prak-
culated more reliably and the results can be presented ticne izkusnje njihove uporabe v Laboratoriju za varnost v
in a graphical form as pictures or animations. How- prometu na Fakulteti za pomorstvo in promet Univerze v Ljub-
ever, attention should be paid in their usage and in in- ljani. Clanek obravnava tudi tehnologijo programskih orodij,
ki omogocajo zajem merodajnih informacij iz fotodokumen-
terpreting the obtained results. Because of the large
tacije cestno-prometne nesrece, ter delo izvedenca ali policista
amount of data needed by these programs, it is almost na kraju cesto-prometne nesreee, ki temelji na tej tehnologiji.
impossible for two reconstruction experts to obtain
the same results by the analysis of the same accidents. KUUCNE BESEDE
When using these programs, the reconstruction expert
must know the physical basis on which the programs cestno-prometna nesreca, preiskovanje, rekonstrnkcija, simu-
work. Blind confidence in beautiful animation without lacija, varnost.
physical background can be very dangerous. As an il-
lustration of the limited use of these programs, only LITERATURE
one simple fact has to be mentioned. If the vehicles
reach the speeds and conditions in which decomposi- (1] G. Genta. Motor Vehicle Dynamics. Modeling and simu-
tion occurs, then the results produced by the computer lation. World Scientific Publishing. 1999
programs have no practical value. (2] S. Fenton, R. Kerr. Accident Scene Diagramming
Using New Photogrammetric Technique. Knott Labo-
The primary advantage of the rectified picture is
ratory, Inc., SAE Technical Papers 970944.
the fact that after being traced a plan view drawing
[3] J, F. KerkhotT. Photographic Techniques for Accident
might be created with geometry of details preserved so Reconstrnction. JFK Engineering. SAE Technical Pa-
that every dimension or distance relevant to the analy- pers 850248.
sis and reconstruction of an accident can be easily ob- [4) J. Kinney, B. Magedanz. TRANS4- a traffic accident
tained from it. The latter in many ways simplifies the photogrammetric system, description of the system and its
work of the reconstruction expert either in the field (at inherent errors. SAE Technical Papers 861417.
the scene of the accident) or in collecting suitable in- (5] J, Kinney, B. Magedanz. TRANS4- a traffic accident
put data needed for credible and accurate (as much as photogrammetric system, description of the system and its
possible) analysis of the causes of road accidents. inherent errors. SAE Technical Papers 861417.

74 Promet- Traffic- Traffico, Vol. 17,2005, No. 2, 65-75

 M. Batista, T. Magister, L. Bogdanovic: Computer Based Road Accident Reconstruction Experiences

[6] R. Limpret, Motor Vehicle Accident Reconstruction and [11] PhotoModeler Pro 4.0, User Manual, Eos Systems Inc.,
Cause Analysis, 5'h Edition, Lexis Publishing, 1999. 2000.
[7] R. McHenry, B. McHenry. McHemy Accident Recon- [12] J. Stannard Baker. Traffic Investigation Manual. The
struction 2000, McHenry Software. Traffic Institute, Northwestern University, 1979.
[8] 1996-2002 McHenry Software, m-smac Input Manual, [13] H. Steffan, A. Moser, The Collision and Trajectory Mo-
Inc, Carry, NC, USA dels for PC-CRASH, SAE Paper 960886.
[9] PC-Crash, Operating Manual, Version 77.0, November [14] P. Swider, J, Unarski, W. Wach, SMASH-Program for
2002, Dr. Steffan Datentechnik, Linz, Austria. Car Accident Simulation, SAE Paper 2000-01-0848.
[10] PC-Rect, Operating & Technical Manual, Version 3.0, [15) Working Model, User's Manual, MSC Software, Cali-
Jan. 2003, Dr. Steffan Datentechnik, Austria. fornia, USA, 2001.

Promet- Traffic- Traffico, Vol. 17, 2005, No. 2, 65-75 75