See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/4005354

Vehicle detection and car speed monitoring

system using GMR magnetic sensors

Conference Paper December 2002

DOI: 10.1109/IECON.2002.1185535 Source: IEEE Xplore

CITATIONS READS
11 892

3 authors:
Jose Pelegri-Sebastia Jorge Alberola
Universitat Politcnica de Valncia Universitat Politcnica de Valncia
42 PUBLICATIONS 277 CITATIONS 10 PUBLICATIONS 107 CITATIONS

SEE PROFILE SEE PROFILE

Vicente Llario
Universitat Politcnica de Valncia
6 PUBLICATIONS 13 CITATIONS

SEE PROFILE

All content following this page was uploaded by Jorge Alberola on 30 July 2015.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
 Vehicle Detection and Car Speed Monitoring
System using GMR Magnetic Sensors
J.Pelegr, J. Alberola, V. Llario
Escola Pol. Sup. Gandia (Polytechnic University of Valencia)
Department of Electronic Engineering
C/ Nazaret-Oliva s/n, 46730-Grao de Ganda (Valencia-
Spain)
E-mail: [email protected]

Abstract A new simple circuit can detect the magnetic Oe) and small magnetic elds (see gure 2).The dieren-
perturbation caused by the cars with two GMR Magnetic
Field Gradient Sensors located on the pavement of a tial output signal of bridge, with respect to time, varies
high- way or fast road without wires. When a car passes accord with the perturbed gradient eld. The ability to
above the board, a Microcontroller processes the signal of detect only magnetic gradients allow low sensitivity to ex-
sensors to obtain the speed and length of the car in real
time. The dates of more than one thousand vehicles could ternal sources of constant magnetic noise. The output of
be stored in ROM until it is full. At this point the sensor is a lineal dierential voltage in 0 to 200 Oersted
microcontroller sends the dates through a FSK modulation, range with a sensitivity of 225 V/Oe (see gure 3).
that a PC stored per- manently, and after present it with a
software. The aord- able system could be used to
automatic trac measurements and control at many places,
replacing the expensive systems like the ultra sonic
sensors and the video cameras to cover an extended area.
Keywords Magnetic eld measurement, GMR sensor, ve-
hicle detection, car speed monitoring.

I. Introduction
Velocity is the cause of most number of trac acci-
dents on highways or fast roads. Our system could pre-
vent it taken advantage of the Magnetoresistive Giant
eect (GMR) of sensors. Reference [1] are an introduction
of this eect and the reference [2][3][4] presents some
applications of the GMR sensors. Fig. 2. Functional block diagram of the GMR sensor.
When a vehicle passes above the board, in a constant
magnetic eld presence, it disturbs the eld. The earths
eld provides a uniform magnetic eld over a wide area
(several Kilometres), and gure 1 shows how a ferrous ob-
ject, a car, creates a local disturbance in this eld whether
it is moving. GMR sensors can detect the change in the
earths eld due to the vehicle disturbance in many types
of applications [5][6][7]. The two GMR sensors used are
the AB001-02 of NVE Corporation. Functional Block of

Fig. 3. Output of sensor is a lineal dierential voltage in 0 to

200
Oersted range [8].
Fig. 1. Ferrous object disturbance in earths eld.

sensors can be represented by a Wheastone Bridge where II. Design of the Board
two pairs of unshielded GMR bridge resistors provide for The system is composed of two GMR sensors (AB001-
directional sensing of small gradients in large (until 250 02), a Microcontroller PIC family, a fast semiconductor
IC memory with an access and write cycle of 45 ns and a
0-7803-7474-6/02/$17.00 2002 IEEE
radio transmitter module with a two kilometer of range, and it can continuously works during several days in an au-
that makes the prototype available with a low cost. tomatic way.
The diagram of blocks is the following one in gure 4. We
can note a GMR sensor, a high pass lter, the instrumen-
tation amplier, a low pass lter. As well a microcontroller
with a A/D converter, a fast ram and the radio module.

Fig. 6. Sensors output signals under low and high current level
excitation.

III. Experimental results

When a car passes through above the sensors in its sensi-
tivity direction we obtain one sensor signal (see gure 7) in
Fig. 4. Diagram of blocks of the electronic circuit. a digital oscilloscope. If we see the under side structure of a
car we can determine the reason of three peaks appeared in
Figure 5 shows the placement of car sensing system, the photo. First peak is due to forward transmission axis of
where we have the two sensors separated a well-known dis- the car, second is the collector of exhaust pipe and third is
tance between theirs (20 cm) for detect the velocity. The caused by the rear transmission axis. The movement of car
two GMR sensors are Magnetic Field Gradient sensors. causes the perturbation of ux lines the constant Earths
eld in both senses, to favour and in against of movement,
and it is the uctuation observed in the signal. The fol-

Fig. 5. The placement of car sensing system.

Fig. 7. Output signal of the sensor captured for one Opel Omega at
We can see the rst prototype in gure 6 The sensors 60 Km/h.
detect the eld variation from the car body when a car
passes through above the board. A Microcontroller cap- lowing image (see gure 8) correspond to another type of
tures the signal with an A/D conversion and processes the car, a Rover at the same speed. We can appreciate that the
signal with a fast memory buer (RAM). Results of num- shape of signal is similar than the Omega signal. The dif-
ber, speed, and length of cars are stored in ROM, prepared ferences between signals can give us a lot of information to
to transfer the data with a radio transmitter, to a PC. recognize the type of car but we will need sensors in three
When transmission is actived by prototype, the computer dimensions. The waveforms of sensor 2 arrives a little in-
shows the information with two graphics in a front panel terval of time retarded respect of the waveforms of sensor
where each one represents the speed and length of each 1. This time increment is taken advantage to calculate the
counted vehicle. A meter represents the average speed of speed of cars by the equation 1,
the highway where is located the board and the interval s
V = (1)
time of the last data send has happened. The system is t
autonomous with 8 units of small battery, because of low where s is the space between the two sensors, exactly 20
power consumption of GMR sensors and of Microcontroller, cm, and t is the time interval between two
waveforms,
 The period of sample rate selected of A/D converter is
1.2 ms. A program code debug reduce the necessary pro-
cessing time among the step of two followed cars. The mi-
crocontroller works at frequency of 32 MHz and it executes
each instruction in 125 ns. The calculation time of velocity
for 128 samples is 625 ms approximately. That mean the
computing of velocity car will be correct if interval of time
between two cars is mayor or equal to one second.
V. Conclusions
GMR sensors can be a good election to get multiple in-
formation about trac event in extremely conditions and
Fig. 8. Output signal of the sensor captured for one Rover 214 at 60
Km/h. with noise presence. The compound system with this sen-
sors have a good stability, with respect the temperature,
and immunity in front of any kind of ambient noise. The
and V is the speed of car in m/s. The length of the car be nal result is a cheap system with a very low installation
estimated by the equation 2 and maintenance cost that it could be adapted to many
applications like the trains circulation control or ferromag-
L = V T (2)
netic pieces movement in industry.
where T is the width of the output waveform. It could be possible to recognize the kind of vehicle with
the placement of more oriented sensors in the highway to
IV. Signal Processing Method obtain a three dimensional signals. A complex software
In order to detect the time interval t automatically, we could analyze the signals to compare it with stored pat-
use the discrete cross correlation (see equation 3) terns.
+ References
y[m] = x1 [n] x2 [n m] (3) [1] Rob ert L.White, Giant Magnetoresistance: A primer.IEEE trans-
actions on Magnetics, vol-28, n 5, June 1992.
n=
[2] James M. Daughton, GMR and SDT Sensor Applications, IEEE
where x1 [n] and x2 [n] are the output digitized voltage of Trans. On Magnetics, V.36, n 5, pp.2773-2778, Sept. 2000.
[3] J. Pelegr, D. Ramrez, E. Sanchis, A.E. Navarro and S.
sensor 1 and sensor 2, respectively. Implement this equa- Casans, Giant magnetoresistive sensor in Conductance control of
tion in the microcontroller supposes to program a code or Switch- ing Regulators,IEEE Trans. On Magnetics, V.36, n 5,
p3578-3580, Sept. 2000.
algorithm so ecient and fast. The system architecture [4] K.-M.H.Lemssen et al, Robust giant magnetoresistance sensors.
used is a microcontroller of 8 bits and A/D converter inte- Sensors and Actuators A-85,Number 1-8, 2000.
[5] T. Uchiyama, K. Mohri, H. Itho, K. Nakashima, J. Ohuchi and
grate on the chip, a parallel ram memory (2K x 8 bit) for Y. Sudo, Car Trac Monitoring System Using MI Sensor Built-
fast processing algorithm. In Disk Set on the Road. IEEE Trans. Magnetics. vol.36, n 5,
pp.3670-3672, Septemb er 2000.
The following routine corr.c computes equation 3 for 0 [6] M. Caruso, Application Notes AN-201, 202, 204 and 207.
k M , with any M N 1. Honewell SSEC, www.ssec.honeywell.com
[7] Pavel Ripka, Magnetic sensors for trac control, International
/* corr.c -sample sample cross correlation */ Symposium on Measurement and Control in Robotics (ISMCR99),
Osaka- Japan, June-1999.
/* of two length-N signals */ [8] Data Sheet Magnetic sensor AB001-02 of NVE Corporation.

void corr ( N, x1, x2, M, y )

double *x1, *x2, *y;
int N, M;
{

int k,n;
 for( k=0; k<=M ;k++ )
for( y[k]=0, n=0; n<N-k; n++ )
y[k] += x1[n+k] * x2[n] / N;
}
Actually, the routine computes the more general sam-
ple cross correlation between two length-N signal blocs
x1 (n), x2 (n), n=0, 1,. . . ,N-1. Where the compute is
y[k], k=0, 1,. . . ,M, the x1 , x2 are N-dimensional (signals),
y is (M+1)-dimensional (result), M is result length (2N+1)
and N is number of samples of x1 , x2 .

View publication stats