You are on page 1of 10



Peggy SUBIRATS (corresponding author)

Centre dtudes Techniques de lquipement Normandie Centre ERA 34
10 chemin de la Poudrire, BP 245, 76121 Le Grand Quevilly Cedex
Email :
Phone : +33 (2) 35 68 81 79
Technople du Madrillet, Avenue Galile BP 10024, 76801 Saint Etienne du Rouvray Cedex
Email :
Phone : +33 (2) 32 91 58 14
Centre dtudes Techniques de lquipement Normandie Centre - ERA 34
10 chemin de la Poudrire, BP 245, 76121 Le Grand Quevilly Cedex
Email :
Phone : +33 (2) 35 68 81 33
Centre dtudes Techniques de lquipement Normandie Centre ERA 34
10 chemin de la Poudrire, BP 245, 76121 Le Grand Quevilly Cedex
Email :
Phone : +33 (2) 35 68 81 34
Centre dtudes Techniques de lquipement Normandie Centre ERA 34
10 chemin de la Poudrire, BP 245, 76121 Le Grand Quevilly Cedex
Email :
Phone : +33 (2) 35 68 81 30

In this paper, we present a tool to evaluate the safety of crossroads. To accomplish this task,
the proposed system detects and registers the near miss accidents. The final goal is to assess
the impact of improvement on existing road designs and crossroads facilities.
This tool uses traffic analyzer like speed radars and pneumatic tubes to detect vehicles on
the main road and its branches. Data coming from the radars and the road tubes are
transmitted through the wireless networks to a processing unit. Then, an algorithm
calculates the time to collision to detect near miss accidents. Finally, the uniqueness of the
proposed solution lies in a video recorder which continuously records and preserves thirty
seconds before and fifteen seconds after the incident. This allows us having a qualitative
analysis, a complete mathematical formulation and easily determining if the near miss
accidents result from a driver error or a road design problem.
This system isn't bulky and can be easily set up on different kinds of crossroads. In fact, a
very short time (nearly one hour) is required to install all its components.

Subirats, Dupuis, Violette, Doucet and Dupre

Rural roads are more dangerous than other roads. In Europe, more than 80% of all fatal
collisions occur on rural roads. Rural roads account for 60% of road fatalities (1, 2) compared
to 10% for motorways (1, 3). Three accident types are reported under car collisions on rural
roads: single-vehicle accidents (e.g., run-off-road with further collisions with dividers and
utility poles); head-on collisions; and collisions at intersections. On the rural roads, one
accident out of 10 occurs at intersections. Their severity is 5.3 times higher than in urban
Facing this reality, we decided to develop a tool to evaluate the safety level of a crossroad.
The objective is to give an indicator of intersection dangerousness. This safety indicator is
defined at the end of this paper.
Here, we first present the context and the needs that motivated the development of this
system. In the second part, we detail the system specifications. Its set up on the intersection is
presented in the third part. In the forth part, we define a safety indicator in a crossroad. In a
last part, we show some results and finally, we conclude and give some perspectives of this


The present study is a part of the French national multidisciplinary research project PREDITSARI. This national project aims at informing drivers and road managers more effectively,
about the high risk of loosing control on the rural road network.
In this context, we have developed a system to fulfil the criteria and the needs mentioned

1.1 A strong challenge

In France, intersections represent less than 1% of the distance traveled by users, but over 10%
of accidents. The risk of an accidents, for a user traveling on a rural road, is multiplied by 10
at an intersection.
1.2 A road manager s need
The danger posed by intersections, results in a strong social pressure to secure these sites. The
user facility requires a safety diagnosis whose goal is to understand accidents process.
However, at intersections, there are only few injury accidents, and often the police report is
incomplete. Indeed, the hearing of the user driving on the secondary road is not completely
take into account because of the lesser degree of seriousness of his injuries. Consequently,
road managers have a lack of data to understand the problems of the site.

Subirats, Dupuis, Violette, Doucet and Dupre

Road managers want to quickly measure the effectiveness of a modified facility, improvement
of facilities or environmental modification. On this type of site, accident analysis requires a
decline of 3 to 5 years to determine the effectiveness of a change. This delay doesnt
correspond to the expectations of managers, or those of users.
To provide a high level security, managers are implementing many innovative developments
such as: crossroads chicane wall effect, dynamic warning signs, rural roundabout and so on.
Managers need to quickly evaluate these innovations.
1.3 Definition of a crossing at grade near-miss and system goals
The system described hereafter intends to understand disorders in a crossroad and to evaluate
the effectiveness of facilities.
It detects and records conflicts between users on the non-priority road who cross the road in
front of the users driving on the priority road in straight motion. This type of conflict
represents the main type of accidents called crossing at grade accidents.
A precise study of the patterns encountered at accident-prone intersections on the rural roads,
showed that some situations are much more dangerous and more common than the others. In
particular, cross movements at the main road cause most of the accidents at the intersections.
These movements are called 1st and 2nd crossing at grade. A crossing at grade occurs when
a vehicle situated on the minor road enters into the intersection and a vehicle driving on the
main road crashes it. This is called a 1st crossing at grade when the vehicle driving on the
main road comes from the left with respect to the vehicle situated on the minor road (Fig 1.a).
This is called a 2nd crossing at grade when the vehicle driving on the main road comes from
the right with respect to the vehicle situated on the minor road (Fig 1.b).



FIGURE 1 Illustration of crossing at grade accident: (a) a 1st crossing at grade, (b) a 2nd crossing at
grade accident

So, to meet the diagnosis and assessment needs, we thought to detect and record near-miss
accidents (and not accidents) related to the movement of 1st and 2nd crossing at grade. This
min goal is to give to the road managers an idea of the safety level of the intersection without
waiting for many accidents to occur.
Thus, the system detects situations that are similar to the concept of traffic conflict defined by
Amunndson in 1977. A traffic conflict (also called near-miss accident) is "an observable
situation, during which two drivers approach one another in time and space, to a point where
there is a risk of accident if their movements remain unchanged " (11,12,13).
Thus, our system detects situations as mentioned above in (11).

Subirats, Dupuis, Violette, Doucet and Dupre


In this part, we present the specifications chosen for developing our system.
2.1 Principle
The proposed system principle is shown in the figure 2.

FIGURE 2 Principle of the near miss accident detection system

Once a non-priority vehicle, i.e, a vehicle coming from the minor road, starts and gets into the
intersection, two confidence intervals are defined (1st junction confidence interval and 2nd
junction confidence interval). If a vehicle driving on the main road is detected during these
intervals, we face a near-miss accident. Both confidence intervals are empirically defined
depending on the kind of intersection. These parameters can easily be changed prior to the
system installation.
The time to collision (TTC) between a vehicle getting into the intersection from the minor
road and a vehicle driving on the primary road is defined as the difference in the time between
the moment when the non-priority vehicle restarts from the stop line and the moment when a
vehicle is detected on the main road.
The time to collision measurement between a minor road driver and a primary road driver is
possibly a representative of an intersection security level.
Furthermore, the addition of a video system in the analysis allows better understanding of why
these situations occurred. By analyzing the videos, we have the possibility to checking if they
result from a driver error or an intersection design problem.
2.2 Technical specifications
In order to be operational and easy to use, the developed system has to satisfy the following
o Detection of vehicle entering the intersection from the minor road,

Subirats, Dupuis, Violette, Doucet and Dupre

o Detection of straight moving vehicles and discard turning vehicle from the major
o Calculation of the time to collision,
o The system has to be installed onto different kinds of intersections,
o The system must use wireless technologies,
o Rapid installation of the system on the intersection without disrupting the traffic,
o The system must work with an autonomous energy like solar energy or windmill.
Measurement taken for one week are enough to have an idea of the intersection safety,
o The system must satisfy the regulations concerning the safety of the users and
roadside obstacles,
o The system should ensure the security against theft (atmospheric condition, etc).


3.1 The system developed
The principle of the system is presented in figure 3.

Priority vehicle

vehicle detection

WIFI transmission

Radio transmission



FIGURE 3 Near-miss accident detection system diagram

First, the system detects vehicle driving on the main roads. Out of many existing traffic
sensors, we chose to use speeds radar. The speed parameter allows to distinguish between
vehicles going straight with the vehicles having a turning movement that cant be involved in
a near-miss accident. As we cant possibly use wired solution, we propose a WIFI
transmission. For this, speed radars send the data measured via a WIFI transmitter to an access
point connected to the central system.
Then, the system detects vehicles situated on the minor road and entering the intersection. For
this, we use a pneumatic tube placed on the stop line. This technology is enough to detect
vehicle and is easy to install. The information about the presence of a non-priority vehicle on
the minor road is sent via a radio transmitter to a radio receiver which is connected to the
central system.
The central system consists of an industrial computer that calculates the time to collision. If
this time is less than 5 seconds (or 7 seconds), we are confronted with a 1st line near-miss
accident ( 2nd line near-miss accident respectively). The scene is continuously recorded and
maintains constant buffer of 30 seconds which can be retrieved when required. Once a near-

Subirats, Dupuis, Violette, Doucet and Dupre

miss accident is detected, a video system records the video of 30 seconds before and 15
seconds after the near-miss accident.
The video allows to understand what happens and if the road infrastructure or the driver
behavior is responsible for the situation.
3.2. Results
Our system was evaluated on a rural road located in the city of Gouy (Normandy, France). The
intersection is illustrated in the figure 4.

FIGURE 4 Example of instrumented intersection

The instrumentation is illustrated in figure 5. On the right down, a box containing radar and
WIFI transmitter is situated on the shoulder. On the right up, a pneumatic tube situated on the
stop line and a box containing the radio emitter.



FIGURE 5 Near-miss accident detection system illustration. (a) the control unit associated with a
registration system. (b up) the system for detecting non-priority vehicles. (b down) the case containing the
system used to detect vehicles on the major road.

Subirats, Dupuis, Violette, Doucet and Dupre

The system stores all vehicle speeds, restarting of non-priority vehicles and near-miss accident
An example of near-miss accident recorded by our system is given in figure 6. The nonpriority vehicle has restarted from the minor road in order to enter the intersection before a
tractor (as we can see in Fig. 6.b). Yet, the vehicle driving on the main road had to apply brake
to avoid the collision.




FIGURE 6 Example of near-miss accident detected with our system


In this part, we propose the definition of a crossroad risk indicator. The proposed indicator is
defined below:

I r = Gi (Ci )

Where Ir is the risk indicator, Gi is the conflict number i and Gi(Ci) is the gravity of the
conflict i.
So,the proposed indicator is equal to the number of conflicts (near-misses), weighted by the
severity of the conflict recorded in a given period (15).
We chose to take the number of conflicts per hour.
The weight of a conflict is defined as followed: the severity of a conflict depends on vehicle
speed and time to collision.
G = KS 2


T = time to collision between both vehicles at the conflict point

S = priority user speed.
K = constant number

Subirats, Dupuis, Violette, Doucet and Dupre


In this paper, we have presented a tool to the evaluate safety of crossroads. This tool can
easily be installed on different kinds of intersections. The developed system detects and
records near-miss accidents.
In perspectives of this work, the system has to be tested on several intersections in order to
validate our empirically defined risk indicator.
Finally, on the long run, our system will be helpful for road managers to complete security

The research presented in this paper was performed as part of the French National project
PREDIT-SARI, sub-them RADARR sponsored by the National Research Agency.

Subirats, Dupuis, Violette, Doucet and Dupre



(1) IRTAD (International Traffic Safety Data and Analysis Group). Fatalities by Road
Location, 2007. Acceded Jul. 1, 2008.
(2) ONSIR.
2008. Acceded
Jan. 16, 2008.
(3) SafetyNet. Annual Statistical Report 2006, 2007. (SafetyNet: EC-Contact no.: TREN-04FP6TRSI2.395465/506723, Project Number: 1.3.2). 1 3 ASR 2006 final.pdf. Acceded Mar. 14,
(4) Violette, M. Voisin (2005), Projet RADARR : Analyse des trajectoires en carrefour ,
CETE Normandie-Centre
(5) F. Conche and All (2009), Scurit des carrefours plans sur routes principales, LCPC,
Opration de recherche Risques Routiers
(6) AIPCR (2003), Manuel de scurit routire .
(7) SETRA, CETE Normandie-Centre (2001), Typologie des accidents en intersection sur
routes interurbaines .
(8) ONISR (2009), La scurit routire en France : bilan de lanne 2008 , La
Documentation Franaise.
(9) N. Muhlrad. (1988), Technique des conflits de trafic. Manuel de lutilisateur Synthse
INRETS n 11.
(10) Picado- Santos. Rodrigues. (1998), The Application of Traffic Conflict techniques in
validation of low cost safety Improvement Measures .
(11) Amundson, F. et Hyden, C. Proceeding (1977)First workshop on traffic conflits.
Institute of Transport Economics Oslo Norway.
(12) Glauz, W et Migletz D. (1980), Application of traffic conflits analysis at intersections
. NCHRP report 219. Transport Research Board. Whashington DC
(13) Muhlrad N (1993), Traffic conflict techniques and other forms of behaviour analysis:
Application to safety diagnoses . 6 th ICTCT Workshop. Salzburg
(14) T Brenac . (1992), Scurit des Routes et des Rues SETRA CETUR
(15) T Brenac (1994), Accidents en carrefour sur routes nationales. Modlisation du nombre
daccidents prdictible sur un carrefour et exemples dapplications . INRETS rapport 185
(16) CETE NC SETRA (2001), Typologies des accidents en intersections sur routes
interurbaines .