Sensors-16-00078.pdf

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/291343125

Dynamic Vehicle Detection via the Use of Magnetic Field Sensors Article  in  Sensors · January 2016 DOI: 10.3390/s16010078

CITATIONS

READS

17

303

6 authors, including: Vytautas Markevicius

D. Navikas

Kaunas University of Technology

Kaunas University of Technology

40 PUBLICATIONS   120 CITATIONS   

40 PUBLICATIONS   90 CITATIONS   

SEE PROFILE

SEE PROFILE

Mindaugas Zilys

Darius Andriukaitis

Kaunas University of Technology

Kaunas University of Technology

27 PUBLICATIONS   111 CITATIONS   

40 PUBLICATIONS   105 CITATIONS   

SEE PROFILE

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Smart traffic monitoring system View project

COST Action CA15225 - Fractional-order systems - analysis, synthesis and their importance for future design View project

All content following this page was uploaded by Mindaugas Cepenas on 23 January 2016. The user has requested enhancement of the downloaded file.

sensors Article

Dynamic Vehicle Detection via the Use of Magnetic Field Sensors Vytautas Markevicius *, Dangirutis Navikas, Mindaugas Zilys, Darius Andriukaitis, Algimantas Valinevicius and Mindaugas Cepenas Received: 12 November 2015; Accepted: 4 January 2016; Published: 19 January 2016 Academic Editor: Andreas Hütten Department of Electronics Engineering, Kaunas University of Technology, Studentu St. 50–418, LT-51368 Kaunas, Lithuania; [email protected] (D.N.); [email protected] (M.Z.); [email protected] (D.A.); [email protected] (A.V.); [email protected] (M.C.) * Correspondence: [email protected]; Tel.: +370-37-300-522

Abstract: The vehicle detection process plays the key role in determining the success of intelligent transport management system solutions. The measurement of distortions of the Earth’s magnetic field using magnetic field sensors served as the basis for designing a solution aimed at vehicle detection. In accordance with the results obtained from research into process modeling and experimentally testing all the relevant hypotheses an algorithm for vehicle detection using the state criteria was proposed. Aiming to evaluate all of the possibilities, as well as pros and cons of the use of anisotropic magnetoresistance (AMR) sensors in the transport flow control process, we have performed a series of experiments with various vehicles (or different series) from several car manufacturers. A comparison of 12 selected methods, based on either the process of determining the peak signal values and their concurrence in time whilst calculating the delay, or by measuring the cross-correlation of these signals, was carried out. It was established that the relative error can be minimized via the Z component cross-correlation and Kz criterion cross-correlation methods. The average relative error of vehicle speed determination in the best case did not exceed 1.5% when the distance between sensors was set to 2 m. Keywords: magnetic field; AMR sensors; vehicle speed detection

1. Introduction Intelligent transport [1], smart cities [2,3] and the Internet of Things [4] are terms that surround all of us nowadays. One of the key factors allowing one to consider intelligent solutions [3–5] specifically designed for urban areas is the constantly decreasing price of sensors capable of providing information about traffic flows, environmental conditions and other parameters. The processes of integrating sensors into the existing infrastructure and forming large sensor networks has thus become increasingly less complicated. In turn, the physical medium is less disrupted and society gains access to a brand new quality of life, facilitated by environmentally responsive public services, continual surveillance and a direct connection with other information users [6]. The capability to detect vehicles is the main element in intelligent transport management systems, allowing us to gather information about the traffic intensity and vehicle speed. Amongst the most popular transport sensors are induction loops, commonly used due to their low price and relatively high signal quality. However, the process of installing such devices normally results in significant damage to an extensive area of the road surface and the loop itself may suffer from the thermal expansion of the road pavement. Anisotropic magnetoresistance (AMR) sensors are an alternative to induction loops capable of measuring even extremely weak magnetic fields. When a vehicle is above the sensor, its metal construction distorts the Earth’s magnetic field and this allows one to determine the location of the vehicle. Sensors 2016, 16, 78; doi:10.3390/s16010078

www.mdpi.com/journal/sensors

Sensors 2016, 16, 78

2 of 9

Magnetoresistive sensors are small, so their installation and maintenance are much less complex and pricey in comparison to induction loops. Nevertheless, since their introduction, they have been subject to little scrutiny or testing in this particular application area. As the Earth’s magnetic field distortion signal can be used not only for the detection, but also for the classification and recognition of transport vehicles [7], it is highly important to ensure sensor effectiveness and the repetition rate of the recognized signal. Various objects and processes taking place in the surrounding environment have a major impact on the readings of magnetic sensors. The factors determining the success of locating a vehicle are multiple, therefore, distinguishing the distortions of the magnetic field caused by the vehicle from those caused by certain environmental changes is a rather difficult task [8,9]. The accuracy of locating vehicles in the environment with no undesirable objects affecting the magnetic field would be significantly higher, but, unfortunately, the possibilities of creating such an environment are slim to none at all. Taking that into account, in comparison to road induction loops the process of locating vehicles using magnetic fields is way more complicated. When conducting the task of locating a vehicle using the magnetic field method, these external factors must be eliminated. The literature analysis reveals that the majority of research in the field of AMR sensors has been carried out in laboratories, which indicates the lack of practical research activity when it comes to real transport vehicles and road environment conditions [10]. 2. Related Works Previous research papers mostly addressed the problems related to the detection of vehicles parked on parking lots or in car parks [11]. The measurement system based on AMR sensors was thus created for the purpose of locating vehicles using the Earth’s magnetic field distortion method. It comprises magnetoresistive sensors produced by four different manufacturers and researchers have focused on analyzing the temperature stability and parameter spread influence on detection relevant to reliability [12]. Following a lengthy period of observation, the researchers determined several shortcomings of the method, which were as follows: dependence of the sensor signal on temperature, residual magnetic moment and diffusion of the sensor parameters during the manufacturing process [13]. Upon analyzing the impact of the vehicle construction on the magnetic field changes in the parameters of the mounted sensors, it was determined that said construction plays a vital role in terms of magnetic field distortion and sensor output signals. Even vehicles of similar construction impact the magnetic field components differently when entering and exiting a parking space. The amplitude and form of an AMR sensor signal is also dependent on how the sensors and vehicles are arranged in terms of their positions magnetic fieldwise [13]. When a vehicle is on top of the sensor, in some cases one can determine the critical zones which display no change in the magnetic field. Based on modeling and practical research results, researchers have proposed a location algorithm reliant on the state criteria. It has been established that the most accurate way of determining the state of the parking location sensors relies on the use of complex criteria [14]. Via the use of this method the researchers observed a 93.3% detection effectiveness. In the case of moving vehicles, the majority of the proposed stationary vehicle detection methods have been proven unsuitable due to the need to not only record the fact of the vehicle’s presence, but also measure its speed and the dependence of the magnetic field distortions on the vehicle construction. It is noteworthy that the majority of publications contain algorithms and methods designed to locate vehicles and determine their speed via the use of AMR sensors, yet most of these papers lack any consideration of the fact that small sensors can be passed by vehicles following different trajectories, which can significantly complicate the task of determining the vehicle type. The bottom of the vehicle under the engine is commonly magnetically open whilst the engine section holds a generator and the ignition starter, each of which generate their own operating magnetic fields. In between the engine department and vehicle interior there is typically a steel (magnetic) partition, similar to the ones found between multiple interior and baggage compartments. Much of uncertainty is caused by the aforementioned partitions, knowing that regardless of their distance to

Sensors 2016, 16, 78 Sensors2016, 2016,16, 16,7878 Sensors

3 of 10 3 of 3 of 109

distance to the AMR sensor, they can act as a vehicle driving nearby, or the one being located above, distance to the sensor, they can act a vehicle driving nearby, or to thea one being located above,of i.e., sum ofAMR thethey magnetic field itsas separate components, partition the vicinity the the AMR sensor, can act as and a vehicle driving nearby, or subject the one being locatedinabove, i.e., the i.e., the sum of the magnetic field and its separate components, subject tonotion a partition in vicinity of the sensors, increase or decrease. Therefore, thethe Z component summagneto-resistive of the magnetic field andcan its separate components, subject tothe a partitionthat in the vicinity of the the magneto-resistive sensors, can increase or decrease. Therefore, the notion that the Z component of the Earth’s magnetic field increases under the vehicle is not necessarily true in all cases, and this magneto-resistive sensors, can increase or decrease. Therefore, the notion that the Z component of ofcan thesignificantly Earth’s magnetic field increases under the vehicle is not necessarily true in all cases, and this complicate the process of identifying located in thein same line asand thethis sensor. the Earth’s magnetic field increases under the vehicle aisvehicle not necessarily true all cases, can can significantly complicate the process of identifying a vehicle located in the same line as the sensor. significantly complicate the process of identifying a vehicle located in the same line as the sensor. 3. Experiments 3.3.Experiments Experiments An experiment was conducted with 24 various models of cars produced by different An was conducted with ofof cars by different An experiment experiment was conducted with 24 24 various various models cars produced produced by sensors different manufacturers seeking to evaluate the possibilities as wellmodels as pros and cons of using AMR in manufacturers seeking to evaluate the possibilities as well as pros and cons of using AMR sensors in manufacturers seeking to evaluate the possibilities well as pros and cons of1)using AMR sensors in the control of transport flows. A brand new vehicleas scanning system (Figure was designed, which the transport flows. AAbrand scanning system (Figure 1)1)was which thecontrol control transport flows. brandnew new vehicle scanning system (Figure was designed,Vehicle which allowed usofofto measure the distortion of vehicle the Earth’s magnetic field caused by designed, vehicles. allowed to measure measure the distortion of the Earth’s field caused by measurement vehicles. Vehicle allowed us uswas to the distortion the Earth’s magnetic field caused by vehicles. Vehicle detection detection conducted using twoof sensors located atmagnetic a distance of 30 cm. The was detection was conducted using two sensors located at a distance of 30 cm. The measurement was was conducted using twoalong sensors a distance of 30capturing cm. The measurement was carried carried out every 1 cm thelocated X axis at direction whilst the vehicle position and out the carried every 1the cmX along the X along axis direction whilst capturing therepeated vehicle position andalong the every 1out cm along axis direction whilst the vehicle and the magnetic field magnetic field parameters. Scanning thecapturing direction of the X axis position was every 20 cm magnetic fieldScanning parameters. Scanning alongof the of the X axisevery was every cm alongof parameters. along the direction thedirection X axissensors was repeated 20 cm along the20direction the direction of the Y axis whilst moving both AMR (Figure 2). repeated the of themoving Y axis whilst moving both(Figure AMR sensors (Figure 2). thedirection Y axis whilst both AMR sensors 2).

RS485 RS485

RS485 RS485

I2C I2C I2C I2C

Distance Distance meter meter

MicroMicrocontroller controller MSP430 MSP430

Figure 1. Experiment structure. Figure1.1.Experiment Experimentstructure. structure. Figure 20cm 20cm

Y

Y

Y

AMR AMR

X X

Y

X X

AMR AMR

40cm 40cm

Figure 2. 2. Vehicle Vehicle scanning scanning scheme. scheme. Figure Figure 2. Vehicle scanning scheme.

For the purpose of gathering data a digital distance measuring instrument was used, which For the purpose of gathering data a digital distance measuring instrument was used, which transferred all the data via a RS 485 communication port to a computer where it was then For the all purpose of gathering dataserial a digital distance measuring was used, which transferred the data via a RS 485 serial communication port to ainstrument computer where it was then processed with the specifically designed software. The data acquired from thewhere sensor the RS transferred all the via a RSdesigned 485 serial communication to a from computer it via wasRS then processed with thedata specifically software. The data port acquired the sensor via the 485 485 port. with The process of transferring data and reading theacquired data from the the AMR sensors via RS the485 I2C processed the specifically designed software. The data from sensor via the port. The process of transferring data and reading the data from the AMR sensors via the I2C interface interfaceprocess was carried out using data a low-power MSP430 microcontroller. port. of transferring and reading the data from the AMR sensors via the I2C interface was The carried out using a low-power MSP430 microcontroller. The Earth’s magnetic field distortions were detected on different vehicles. A sample set of the was carried out using a low-power MSP430 microcontroller. The Earth’s magnetic field distortions were detected on different vehicles. A sample set of the Earth’s magnetic field distortions data caused by detected several of them is provided in Figure 3. The research The Earth’s magnetic field distortions were different vehicles. A sample setresearch of the Earth’s magnetic field distortions data caused by several ofon them is provided in Figure 3. The results indicate that the change in the magnetic field modulus and separate components when the Earth’s fieldthe distortions by several of them isand provided in Figure 3. The when research resultsmagnetic indicate that change data in thecaused magnetic field modulus separate components the vehicle is driven on top change of the magneto-resistive sensor is different and depends on the vehicle brand results that in the magnetic field modulus andand separate components when the vehicleindicate is driven on the top of the magneto-resistive sensor is different depends on the vehicle brand vehicle is driven on top of the magneto-resistive sensor is different and depends on the vehicle brand

Sensors 2016, 16, 78 Sensors 2016, 16, 78

4 of 9 4 of 10

andthe theposition positionofofsensors sensorswith withrespect respecttotoa aparticular particular vehicle (Figure This reason behind and vehicle (Figure 3).3). This is is thethe reason behind the the that fact that when a vehicle passes the sensor in different places (with respect to the Y axis), very fact when a vehicle passes the sensor in different places (with respect to the Y axis), very diverse Sensors 2016, 16, 78 4 of 10 diverse magnetic field are profiles are recorded. As the a result, the based methods based on measuring and magnetic field profiles recorded. As a result, methods on measuring and comparing and the position of sensors with respect to a particular vehicle (Figure 3). This is the reason behind comparing absolute values arefor unsuitable detecting the type of and speed of vehicles. absolute values are unsuitable detectingfor the type and speed vehicles. the factcompleting that when athe vehicle passes the sensor different data, places (with the magnetic Y axis), very Upon interpolation of acquired the inacquired fullofrespect view ofto the field Upon completing the interpolation of the data, a full aview the magnetic field distortion diverse magnetic field profiles are recorded. As a result, the methods based on measuring andare distortion caused by a vehicle was acquired (Figure 4). From the data it can be noticed that there caused by a vehicle was acquired (Figure 4). From the data it can be noticed that there are a number comparing absolute values are unsuitable for detecting the type and speed of vehicles. a areas number of areas as “Dead zone”) under the vehicle where the magnetic field remains of (marked as (marked “Dead the zone”) under the where the magnetic remains undistorted Upon completing interpolation ofvehicle the acquired data, a full viewfield of the magnetic field undistorted (as if the vehicle was not present at all). Such situations are critical when using AMR (as if the vehicle was by nota vehicle presentwas at all). Such situations arethe critical sensors distortion caused acquired (Figure 4). From data itwhen can be using noticedAMR that there are for sensors for determining the presence of a vehicle in a static state. determining theofpresence of a vehicle in azone”) static state. a number areas (marked as “Dead under the vehicle where the magnetic field remains

1000

Y1

400

600

800

1000

-200 -600

0 -200

400

600

800

1000

200200

400

600 Z1

800

1000

1200

M ag m etic field , m G

200

400 200

0 200

400

0 200

400

600

800

600 cm X position,

800

400

1000

800

1000

1200

Y2

400

600

800

1000

1200

X position, cm

400

600

Z2 800

1000

1200

X position, cm Z2

800

600

600

400 400

200 200

0 200 0

1200

1000

Y2

600

X position, cm

800

600

1200

-200

M ag m etic field , m G

M ag m etic field , m G

400

1000

0

-400 200

Z1

800

600

800

X position, cm

-200

X position, cm

800

600

0

-400 200

1200

400

200

X position, cm

-600 200

-600 -800

1200

Y1

-400

-800 200

1200

X position, cm

-400 -600 200

M ag m etic field , m G

800

200 M ag m e tic fie ld , m G

-400

600

-4000

X position, cm

0 -200

400

-600 -800 200

200 M ag m e tic fie ld , m G

-400

X2

200 -200

M ag m e tic field , m G

-800 200

0 -200

M ag m e tic fie ld , m G

-600

0

M ag m e tic fie ld , m G

-400

X1

200

-200

M ag m e tic field , m G

0

M ag m e tic field , m G

M ag m e tic field , m G

undistorted (as if the vehicle was not present at all). Such situations are critical when using AMR X1 X2 200sensors for determining the presence of a vehicle in a static 200 state.

1200

200

400

400

X position, cm

600

800

1000

600 X position,800 cm

1000

1200

X position, cm

Figure 3.3.The of magnetic measured by twodistinct distinctsensors sensors (plots ofmagnetic magnetic Figure The distribution of the the field distinct sensors (plots Figure 3.distribution The distribution of the magnetic fieldmeasured measured by by two two (plots of of magnetic field X1, and X2, Y2, Z2 ofZ2 sensors 1 and 2 respectively). PlotPlot curves areare color coded fieldcomponents components X1,Y1, Y1,Z1 Z1 and X2, Y2, Z2 ofofsensors 11 and 22 respectively). respectively). Plotcurves curves are color field components X1, Y1, Z1 and X2, Y2, sensors and color by sensor along latitudinal vehiclevehicle line. coded by position sensor position along latitudinal coded by sensor position along latitudinal vehicleline. line.

VehicleNo. No.11 Vehicle Z1

Z1

Z1

Z1

800

800

700

Dead zone

700

Dead zone

200

Magmetic field, mG

400

600

Magmetic field, mG

600

Magmetic field, mG

Magmetic field, mG

600

400 200 0

0

200

200 Yp

Yp

150 os

iti o n

150 os

iti o n

100 ,c

m

100 ,c

m

50

200

0

50 0

200

400

600

800

800

, cm 600 osition 400 X p m c , n io X posit

1000

1000

600

500

500

400

400

300

300

200

200 100 1000 0 -100 -100

Figure 4. Cont. Figure 4. Cont.

Figure 4. Cont.

0

0

200

200

400 600 X position, cm

400 600 X position, cm

800

800

1000

1000

1200

Sensors 2016, 16, 78

5 of 9

Sensors 2016, 16, 78

5 of 10

Vehicle No 2 Z1

Z1

700

Magmetic field, mG

Magmetic field, mG

600 600 500 400 300 200

200 Yp

150 100 os iti o n, cm

50

400

200

0

600

X positi

800

1000

500

Dead zone

400 300 200

on, cm

100

0

200

400 600 X position, cm

800

1000

800

1000

Vehicle No. 3 Z1

Z1

650 600

Magmetic field, mG

Magmetic field, mG

550 600 500 400 300 200 200 Y p 150 100 os iti o n, cm

50 0

200

400

600

800

ion, cm X posit

1000

500

Dead zone

450 400 350 300 250 200 150 0

200

400 600 X position, cm

FigureFigure 4. The4. distribution of the magnetic field component)caused caused three different The distribution of the magnetic fielddistortion distortion (Z component) byby three different vehicles (different color means different position of sensor with respect Y axis—across the vehicles (different color means different position of sensor with respect to thetoY the axis—across the vehicle). vehicle).

4. Analysis of Vehicle Detection Methods

4. Analysis of Vehicle Detection Methods

The key must must be addressed whenwhen designing the system for detecting speed andand vehicle Thetask keywhich task which be addressed designing the system for detecting speed type isvehicle relatedtype to the accuracy of determining the duration between signals from separate sensors. Once is related to the accuracy of determining the duration between signals from separate established the stable signal discretization andfrequency completing matching both signals, sensors. Once established the stable signalfrequency discretization and the completing the of matching of it becomes possible to calculate the delay between signals and the vehicle both signals, it becomes possible to calculate the delay between signals andspeed. the vehicle speed. Seeking to evaluate the accuracy of determining the vehicle speed using AMR sensors, Seeking to evaluate the accuracy of determining the vehicle speed using AMR sensors, we we turned turned to the data acquired from conducting the experiment and carried out a comparison 12 to the data acquired from conducting the experiment and carried out a comparison of 12 of selected selected methods. The list of these methods is provided in Table 1. methods. The list of these methods is provided in Table 1. Table 1. List of methods.

Table 1. List of methods. No. 1 No. 2 31 4 52 63 74 85 9 6 10 7 11 8 12

Method Z component peak detection Method Z component cross-correlation Module detection Z componentpeak peak detection Module cross-correlation Z component cross-correlation Vectorial deviation peaks (Equation (1)) Modulecross-correlation peak detection(Equation (1)) Vectorial deviation Combined vectorial Module deviation—peaks cross-correlation(Equation (2)) Combined vectorial deviationpeaks cross-correlation (2)) Vectorial deviation (Equation (Equation (1)) Kz criterion peaks Vectorial deviation cross-correlation (Equation (1)) Kz criterion cross-correlation Combined vectorial deviation—peaks (Equation (2)) Z peaks ±50 readings of cross-correlation CombinedModule vectorial deviation cross-correlation (Equation (2)) peaks ±50 readings of cross-correlation

9 10 11 12

Kz criterion peaks Kz criterion cross-correlation Z peaks ˘50 readings of cross-correlation Module peaks ˘50 readings of cross-correlation

Sensors 2016, 16, 78

6 of 10

Sensors The 2016, selected 16, 78

methods are based on either the process of finding the peak signal values or6 their of 9 superposition in time whilst calculating the delay, or via calculation of the cross-correlation of these signals and determining the lag. The selected methods based on either the process Vectorial deviation is a are simple criterion defined as [11]:of finding the peak signal values or their superposition in time whilst calculating the delay, or via calculation of the cross-correlation of these K | cos   cos  0  cos   cos  0  cos  cos  0 | (1) signals and determining the lag. Vectorial deviation between is a simple criterionfield defined asand [11]:x, y, z axis respectively. where α, β, γ—angle magnetic vector

Combined vectorial deviation criterion compiled K “|cosα is ´ acosα ´ cosβfrom cosγ“square” ´ cosγ0 |and vectorial deviations. (1) 0 | ` | cosβ 0 | ` |the According to the fact that the Z component increases when a vehicle is moving over and decreases where α, β, γ—angle between magnetic field vector and x, y, z axis respectively. nearby the vehicle, it can be used to increase the sensor’s sensitivity when a vehicle is moving: Combined vectorial deviation is a criterion compiled from the “square” and vectorial deviations. K Z| cos   cos0 increases  cos  when cos 0a| vehicle  ( Bz / Bis 1) According to the fact that the component over and decreases (2) z 0 moving nearby the vehicle, it can be used to increase the sensor’s sensitivity when a vehicle is moving: where cos α, cos β—cosines of magnetic field vector’s angles on the influence of vehicle, cos α0, cosβ ´ cosβ K “|cosα ´ cosα (2) z {B 0 | ` | without 0 | ` pBof z0 ´ 1q B —magnetic induction cos β0—cosines of magnetic field vector angles influence vehicle, z on thecos influence of vehicle, ofBzmagnetic value without influenceof ofvehicle, vehicle. cos α0 , cos where α, cos β—cosines fieldinduction vector’s angles on the influence 0 —magnetic β0 —cosines of magnetic field vector angles without influence of vehicle, B —magnetic induction on z Kz criterion: the influence of vehicle, Bz0 —magnetic induction value without influence of vehicle. B K z  z 1 Kz criterion: (3) BB z z0 Kz “ ´1 (3) Bz0 where B —magnetic induction value without influence of vehicle. z0

whereThe Bz0 —magnetic inductiondata value without influence vehicle. which are interpreted in search for view of the gathered when using differentofmethods The view of the gathered data when using different methods which are interpreted in search for signal matching, is presented in Figure 5. signal matching, is presented in Figure 5. 900

Magmetic field, mG

800 700 600 500 400 300 200 100 0 0

200

400 1

600 2

Data

3

800 4

1000

1200

5

Figure 5. The gathered data when using different methods (1—Z component, 2—Module, 3—Vectorial deviation, deviation, Figure 5.4—Combined The gatheredvectorial data when using 5—K different methods (1—Z component, 2—Module, 3— z criterium). Vectorial deviation, 4—Combined vectorial deviation, 5—Kz criterium).

In Figure 6 a view of the data samples from two matching sensors via the use of Z component cross-correlation mismatch data samplessensors from the two be due In Figure 6isa presented. view of theThe data samplesbetween from two matching via thesensors use of could Z component tocross-correlation the different sensitivity or positioning of sensors. Such a mismatch could be eliminated by using is presented. The mismatch between data samples from the two sensors could be data techniques. due normalization to the different sensitivity or positioning of sensors. Such a mismatch could be eliminated by

using data normalization techniques.

Sensors Sensors 2016, 2016, 16, 16, 78 78

of10 9 77of

Sensors 2016, 16, 78

7 of 10

300

1 Sensor 2 Sensor

100 0

Magnetic field, mG

Magnetic field, mG

200

-100 -200

300

1 Sensor 2 Sensor

200 100 0 -100

-300 0-200

100

200

-300 0

100

300

200

400

300

500 X position a)

400

500 X position

300

700

600

700

800

900

800

900

1000 1000

a)

1 Sensor 2 Sensor

300

1 Sensor 2 Sensor

100 200 Magnetic field, mG

Magnetic field, mG

200

600

0 100

-100

0

-200 -100 -300 -200 0

100

200

-300 0

100

300

200

400

300

400

500 X position

600

500 b) X position

600

700 700

800

900

800

900

1000 1000

b)

Figure sensors (a)(a) and samples matching (b) (b) via via the the use use of the Figure 6. 6. A A view view of ofthe thedata datasamples samplesfrom fromtwo two sensors and samples matching of Figure 6. A view of the data samples from two sensors (a) and samples matching (b) via the use of Z component. the Z component. the Z component.

Upon conducting the analysis and processing of the data collected during the experiment using Upon conducting the analysis and processing of the data collected during the experiment using Uponsignal conducting the analysis and of the data the experiment using the selected matching methods, weprocessing have determined thecollected averageduring and maximum relative errors the selected signal matching methods, we have determined the average and maximum relative errors the selected signal matching methods, we have determined the average and maximum relative errors when the distance between sensors was considered to be equal to 2 m. The results are displayed when thethe distance between sensors was considered to be equal to 2 m. The results are displayed in when in Table 2. distance between sensors was considered to be equal to 2 m. The results are displayed in Table 2. Table The 2. analysis of the acquired results has revealed that the smallest average relative errors are TheThe analysis of of thetheacquired that the the smallest smallestaverage averagerelative relative errors analysis acquiredresults resultshas has revealed revealed that errors areare obtained with the second and the tenth methods (the maximum errors do not exceed 6.5%, Figure 7). obtained with the second and the tenth methods (the maximum errors do not exceed 6.5%, Figure obtained with the second and the tenth methods (the maximum errors do not exceed 6.5%, Figure 7). 7). The dispersion of errors in the case of using these methods is shown in Figures 8 and 9. TheThe dispersion of of errors ininthe is shown shownininFigures Figures8 8and and dispersion errors thecase caseofofusing usingthese these methods methods is 9. 9. Table 2. Method errors. Table2. 2.Method Method errors. Table errors. Method No. Method No. Relative Error Relative Error (%)(%) Relative Error (%) 1 11 2 22 3 4 5 6 7 88 999 10 33 44 55 66 77 10 11 10 1111 Average 3.6 1.5 2.8 3.3 3.7 6.7 3.3 2.6 3.9 1.5 Average 3.63.6 1.51.5 2.82.8 3.3 3.3 2.6 3.9 1.5 3.1 3.1 Average 3.3 3.7 3.7 6.7 6.7 3.3 2.6 3.9 1.5 3.1 Maximum 28.5 6.0 14.5 15.0 15.0 23.5 Maximum 28.5 6.0 6.0 14.5 14.5 15.0 15.0 15.0 15.0 23.5 23.5 12.5 12.5 12.5 31.0 6.5 15.5 Maximum 28.5 12.5 12.5 31.0 31.0 6.5 6.5 15.5 15.5 16

16

14

14

12

Relative error %

Relative error %

12

10

10 8

8 6

6 4

4 2

2

0 1

0 1

2

2

3

3

4

4

5

5

6 7 8 Method No.

6

7

8

9

9

10

10

11

12

11

12

Method No. Figure 7. Box plot of the methods’ (Table 1) relative errors.

Figure 7. Box plot of the methods’ (Table 1) relative errors.

Figure 7. Box plot of the methods’ (Table 1) relative errors.

12 1212 2.52.5 2.5 11.0 11.0 11.0

Sensors 2016, 16, 78

8 of 10

Sensors 2016,2016, 16, 78 Sensors 16, 78

8 of 10 8 of 9

6 64

Relative error % Relative error %

42 20 0

50

100

150

200

0-2 0

50

100

150

200

-2-4

Distance (cm)

-4-6

Distance (cm)

-6

Figure 8. The dependence of of error position of sensors with respect vehicles the second Figure 8. The dependence errorononthethe position of sensors with to respect tousing vehicles using the method. second method. Figure 8. The dependence of error on the position of sensors with respect to vehicles using the second method. 8

Relative error % Relative error %

6 8 4 6 2 4 0 2

0

50

100

150

200

-2 0

-4 -2 -6 -4

0

50

100

150

200

Distance (cm)

Figure 9. The dependence of error on the position of sensors with respect Distance to vehicles (cm) using the tenth method. -6

Figure 9. The dependence of error on the centerline position of sensors respect to vehicles using8the tenth Sensor placement near the longitudinal of the with passing vehicle (Figures and 9) increases method. the precision of the speed detection. The analysis of the obtained results indicates that if the right Figure 9. The dependence of error the position of sensors with to vehicles using the tenth error) to signal matching method of AMR dataonsamples is selected, it is respect possible (with an acceptable Sensor placement near the longitudinal centerline of the passing vehicle (Figures 8 and 9) method. determine the speed and type of the vehicle (considering the length of vehicle). The benefits of such a increases the precision of the speed detection. The analysis of the obtained results indicates that if the system aresignal as follows: right matching method of AMR data samples is selected, it is possible (with an acceptable

1. 2. 3.

Sensor placement near the longitudinal centerline of the passing vehicle (Figures 8 and 9) error) to the determine and type of the The vehicle (considering theclassification length of vehicle). The benefits increases precision ofspeed the speed detection. analysis of the obtained results indicates that ifaccording the The detection of thethe vehicle location, measurement of speed and of vehicles of such a system are as follows: right signal matching method of AMR data samples is selected, it is possible (with an acceptable to their type. error) toinfrastructure determine and type of the vehicle (considering the length of vehicle). The 1. The detectionthe of speed the vehicle location, measurement of speed and classification of benefits vehicles Lower costs. of such a system are as follows: according to their type.

Relatively modest amount of data to be transferred due to the fact that signal processing can be

2. The Lower infrastructure 1. detection of the costs. vehicle location, measurement of speed and classification of vehicles performed in the sensor itself. 3. according Relativelytomodest amount of data to be transferred due to the fact that signal processing can be their type. performed inResearch the sensor itself. 2. Lower infrastructure costs. 5. Ideas for Further 3. Relatively modest amount of data to be transferred due to the fact that signal processing can be There is great in potential for further research into the use of AMR sensors in the presence of a performed the sensor itself.

different Z component angles (i.e. at different locations across the Earth) and for analyzing different types of signal processing methods for vehicle speed detection. Moreover, there is still a knowledge gap in terms of researching vehicle detection using the magnetic vehicle signature features. 6. Conclusions By experimental research it has been established that AMR sensors can be used for the purposes of vehicle speed control and vehicle classification. Upon conducting research with regard to the different

Sensors 2016, 16, 78

9 of 9

impact of distinct vehicle types on the Earth’s magnetic field it has been established that different vehicles have distinct magnetic signatures which can be used for identifying the vehicle type. According to the research results it can be concluded that is sufficient to use only one AMR Z component for vehicle speed detection. Upon analyzing the 12 methods found suitable for the purpose of speed control, it has been established that the minimum errors are obtained using the Z component for methods 2 and 10 (the Z component cross-correlation and Kz cross-correlation criterion methods). When using the proposed methods the average relative error of speed determinations in the case of applying the most suitable method does not exceed 1.5% when the distance between sensors is 2 m. It should be noted however that the vehicles can affect measurements by adding additional errors. These can be eliminated through other magnetic field components. Author Contributions: V.M. supervised the project direction. V.M., M.C. and A.V. designed the hardware. D.N., D.A. and M.Z. implemented the hardware. M.Z., D.N. and D.A. evaluated the system by developing software. D.N. developed the final software. V.M and M.C. analyzed the data. V.M., D.N. and M.C. wrote the initial manuscript; M.Z., D.A. and A.V. participated actively in the development, revision and proofreading of the final manuscript. Conflicts of Interest: The authors declare no conflict of interest.

References 1. 2. 3. 4. 5. 6. 7.

8. 9. 10.

11. 12. 13. 14.

Schulz, W.; Geis, I. Future role of cost-benefit analysis in intelligent transport system-research. IET Intell. Transp. Syst. 2015, 9, 626–632. [CrossRef] Mone, G. The New Smart Cities. Commun. ACM 2015, 58, 20–21. [CrossRef] Glasmeier, A.; Christopherson, S. Thinking about smart cities. Camb. J. Reg. Econ. Soc. 2015, 8, 3–12. [CrossRef] Lee, I.; Lee, K. The Internet of Things (IoT): Applications, investments, and challenges for enterprises. Bus. Horiz. 2015, 58, 431–440. [CrossRef] Surgailis, T.; Valinevicius, A.; Markevicius, M.; Navikas, D.; Andriukaitis, D. Avoiding Forward Car Collision using Stereo Vision System. Elektron. Elektrotech. 2012, 18, 37–40. [CrossRef] Zhu, H.; Yu, F. A Vehicle Parking Detection Method Based on Correlation of Magnetic Signals. Int. J. Distrib. Sens. Netw. 2015, 1–13. [CrossRef] Cirilo Gimeno, R.V.; Pla Castells, M.; Celda, A.G.; Plume, J.M. A simulation model using AMR sensors applied to vehicle classification and identification. In Proceedings of the ISC’2013, Ghent, Belgium, 13–15 November 2013. Zhou, X. Vehicle location estimation based on a magnetic sensor array. In Proceedings of the 2013 IEEE Sensors Applications Symposium (SAS), Galveston, TX, USA, 19–21 February 2013; pp. 80–83. Marshall, S.V. Vehicle detection using a magnetic field sensor. IEEE Trans. Veh. Technol. 1978, 27, 65–68. [CrossRef] Pitton, A.C.; Vassilev, A.; Charbonnier, S. Vehicle Re-Identification with Several Magnetic Sensors. In Proceedings of the AMAA’2012, Berlin, Germany, 30–31 May 2012; Springer: Berlin, Germany; Heidelberg, Germany; pp. 281–290. Daubaras, A.; Zilys, M. Vehicle Detection based on Magneto-Resistive Magnetic Field Sensor. Elektron. Elektrotech. 2012, 2, 27–32. [CrossRef] Markevicius, V.; Navikas, D. Adaptive Thermo–Compensation of Magneto–Resistive Sensor. Elektron. Elektrotech. 2011, 8, 43–46. [CrossRef] Daubaras, A.; Markevicius, V.; Navikas, D.; Cepenas, M.; Zilys, M.; Andriukaitis, D. Vehicle Influence on the Earth's Magnetic Field Changes. Elektron. Elektrotech. 2014, 4, 27–32. Daubaras, A.; Markevicius, V.; Navikas, D.; Zilys, M. Analysis of magnetic field disturbance curve for vehicle presence detection. Elektron. Elektrotech. 2014, 5, 80–83. [CrossRef] © 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons by Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).

View publication stats