Content uploaded by Mohd Suhaimi Nur Farahana
Author content
All content in this area was uploaded by Mohd Suhaimi Nur Farahana on Dec 05, 2015
Content may be subject to copyright.
Abstract— This paper discussed on automatic parking system
and electronic parking fee collection based on vehicle number
plate recognition. The aim of this research is to develop and
implement an automatic parking system that will increase
convenience and security of the public parking lot as well as
collecting parking fee without hassles of using magnetic card.
The auto parking system will able to have less interaction of
humans and use no magnetic card and its devices. In additions
to that, it has parking guidance system that can show and guide
user towards a parking space. The system used image
processing of recognizing number plates for operation of
parking and billing system. Overall, the systems run with
pre-programmed controller to make minimum human
involvement in parking system and ensure access control in
restricted places. This paper presents algorithm technology
based method for license plate extraction from car images
followed by the segmentation of characters and reorganization
and also develop electronics parking fee collection system based
on number plate information.
Index Terms— Automatic parking system, autocorrelation,
image processing, mean square error, plate number recognition
structural similarity index.
I. INTRODUCTION
The car license plate recognition identification is an
important application in the field of Intelligent Transport
System (ITS) and Electronic toll collection (ETC). The
objective is to extract and recognize vehicle registration
numbers from car images, process the image data finally
utilize for access record and prepare electronic bill.
Electronic toll collection (ETC) or Electronic Car parking
payment is one of the major research topics in intelligent
transportation system (ITS) [1]. ETC is an implementation of
a road pricing concept in order to create benefits such as
increasing the capacity of toll stations, reducing a toll paying
time, enhancing the convenience and safety of travelers, and
minimizing air pollution and fuel consumption. It enables
freeway to parking lot, toll plaza, and bridge, tunnel, and
turnpike operators to save on staffing costs while reducing
delay for travelers and improve overall traffic performance
and parking system. Moreover, monitoring the vehicle traffic
and the management of parking areas are the most
labor-intensive job. Therefore the research on systematic full
automatic parking system is proposed. It differs from
conventional parking system, no magnetic card is used to
record the entry and exit time of. Also, it is designed in such a
Manuscript received February 22, 2012, revised March 8, 2012.
M. M. Rashid, A. Musa, N. Farahana and A. Farhana are with
Mechatronics Eng. Department International Islamic University Malaysia
Kuala Lumpur, Malaysia (email: mahbub96@gmail.com).
M. Ataur Rahman is with the Mechanical Eng. Department International
Islamic University Malaysia Kuala Lumpur, Malaysia.
way that it has the ability of giving out the information
regarding parking free spaces to users before entering the
parking spaces.
In the recent years, License Plate Recognition (LPR) are
having a wide impact in people’s life as their scope is to
improve transportation safety and mobility and to enhance
productivity through the use of advanced technologies [1].
This system is useful in many fields and places such as
parking lots, private and public entrances, border control,
theft and vandalism control. For the past two decades, there
have been various studies for number plate recognition in
images. Early day’s number plate recognition systems
employed detection methods based on techniques such as
edge analysis [1-3], color analysis [4-5], and the Adaboost
training [6]. S.Z. Wang and H.M. Lee [1], D. Zheng et al. [2],
and P. Rattanathammawat and T.H. Chalidabhongse [7]
proposed an edge analysis method for license plate detection.
Ernst [7-9] introduced a face detection method based on local
structure patterns computed by the Modified Census
Transform (MCT).
License plate recognition applies image processing and
character recognition technology to identify vehicles by
automatically reading the license plates. Basically, to build
this system it consists of some major parts which are vehicle
number plate extraction, characters segmentation, and
characters recognition [10-13]. Vehicle number plates
extraction from the car plate images. Before extracting the
number plate, the captured vehicle image should have been
converted into binary format [14]. After extracting the
number plate, the characters are segmented using vertical and
localization on the binary image. Optical Character
Recognition (OCR) algorithm is used to recognize the
character with condition, the background of the image has no
or very little noise [15]. The aim of the system is to recognize
the license plate number of car of the parking place by
algorithmic and introduce magnetic card less parking and
billing activities.
II. SYSTEM OVERVIEW
Recognition of any License Plate Recognition system is
the effectiveness of its algorithms. Six primary algorithms are
used for this License Plate Recognition system.
III. LICENSE PLATE LOCALIZATION
Firstly the car image is captured. Then, the system should
extract the number plate of the car alone for the segmentation
of character purpose. This plate localization algorithm is
based on combining morphological operation sensitive to
specific shapes in the input image with a good threshold
Automatic Parking Management System and Parking Fee
Collection Based on Number Plate Recognition
M. M. Rashid, A. Musa, M. Ataur Rahman, and N. Farahana, A. Farhana
International
Journal of Machine Learning and Computing, Vol. 2, No. 2, April 2012
93
value by which the license plate is located. A big percentage
of localization of License plates is achieved by this algorithm.
This variance can further compound the complexity for an
algorithm to ascertain what area of a vehicle constitutes a
license plate and what area is not. Therefore, the algorithm
must rule out a vehicle's mirror, headlight, bumper etc. In
general, algorithms look for geometric shapes of rectangular
proportion. However, since a vehicle can have many
rectangular objects on it, further algorithms are needed to
validate that the identified object is indeed a license plate. To
accomplish this, key components of the algorithm look for
characteristics that would indicate that the object is a license
plate. The algorithm searches for a similar background color
of unified proportion and contrast as a means to differentiate
objects on a vehicle. Vehicles are moving objects and their
rate of velocity must be accounted for in the algorithm's
design. This speed creates further complexity as a license
plates image is angularly skewed and subjected to refractory
issues from light changes.
IV. LICENSE PLATE SIZING AND ORIENTATION
Components of algorithms that adjust for the angular skew
of the license plate image to accurately sample, correct, and
proportionally recalculate to an optimal size.
Fig. 1. License plate sizing sequence
V. CHARACTERS SEGMENTATION
Character segmentation is an important step in license
plate recognition system. The segmentation of characters in a
license plate is performed by using the following steps.
It is very important for the good performance of character
segmentation. The preprocessing consists of the
determination of plate kind. There are two kinds of license
plate in India. One is black characters in yellow background
and the other is black characters in white background. The
color image is transformed into gray scale image.
Fig. 2. Character segmentation
One of the study use Red color algorithm approach [5].
The techniques will search yellow pixel or some that closer to
yellow in value from the image. Red pixel is set to 1 and
others will be 0. Then, the recognized image will be
converted in binary format for further template matching
approach.
VI. NORMALIZATION
The algorithm in regulating the contrast and brightness of
the captured license plate image is shown below in Fig. 3.
Fig. 3. Character normalization
VII. OPTICAL CHARACTER RECOGNITION (OCR)
This is the process that identifies individual alpha numeric
characters on a license plate. Algorithms also look for
characters of equal color and equidistance, with similar font
structures to break apart each individual character. This
sequential congruency of the characters embodies a
characteristic set that is typically uniform, regardless of the
type of license plate. Character Segmentation separates each
letter or number where it is subsequently processed by optical
character recognition (OCR) algorithms. It translates the
captured image into an alpha numeric text entry.
Fig.4 Optical character recognition process
VIII. SYNTACTICAL /GEOMETRICAL ANALYSIS
Algorithm to verify alpha numeric information and
arrangement with a specific rule set. The algorithms operate
sequentially with instructions being executed in milliseconds.
The successful completion of each algorithm is required
before subsequent algorithms can be operational.
Fig. 5. Syntactical analysis
IX. METHODOLOGY
The system is divided into sub-systems which are ‘FULL’
display system, image acquisition and plate number
recognition, auto direction system and auto payment system.
Firstly, data is acquired from ultrasonic sensors of each
parking space to count the availability of parking spaces in
the parking area. Then, image of the car is acquired in the
entrance to be analyzed. During this time, time entering and
license plate reference number are recorded for future
transaction.
Secondly, the image acquired from hardware components
by Camera are analyzed in data analysis part where mostly
done in Matlab. The image of the car is analyzed for its
license plate numbers for future reference. The time entering
is analyzed during car exiting to calculate the fees of parking.
Plus, the data acquired from ultrasonic probes of each
International Journal of Machine Learning and Computing, Vol. 2, No. 2, April 2012
94
parking space is calculated for the calculating free parking
spaces.
The ‘FULL’ sign is expected to be displayed thru LCD
display. Other information to be displayed is fees
accumulated according to specific parking rate. The fee is
likely displayed on the desktop computer interface. The
direction system also expected to be having information
display where it is used to guide drivers to nearest free
parking space. All the displayed are the output from
MATLAB programming.
In addition, the auto parking system needs interface and
management systems to communicate with drivers and
system’s developers. Firstly, the barrier gate is managed to be
opened when the main program acquired the license plate
reference and it is recorded. And after some time delayed, the
barrier gate will close again. Then, all the output need no
physical interactions will be managed by LCD displays.
However, for software development and calculation, the
desktop computer will be used. System’s developers and
maintenance would have to use the desktop computer to
manage the systems.
Finally, complete content and organizational editing
before formatting. Please take note of the following items
when proofreading spelling and grammar.
X. SCOPE AND LIMITATIONS
Number plate is differed in term of color, size and type
from country to country. Different algorithm has to be
applied for different type of number plates. As for Malaysia
number plate; the method used based on the morphological
algorithms and connected components analysis, including six
major stages, which are, RGB to binary conversion, image
filtration, analysis and dilation, extracting the accurate
location of the license plate and characters recognition.
Although the proposed method is designed particularly for
Malaysian license plates, it can be readily extended to cope
with license plates of other countries, especially those using
Latin characters. In general, Malaysian car plates are
classified in several various categories in terms of format and
colors. 3 major types are; (1) personal vehicles license plates
which are composed of a black background with white
characters organized into single or double row, (2) taxis
license plates are composed of a white background with
mostly a single row of black characters, and (3) special types
of license plates such as Proton, Satria, Sukom, Tiara and etc.
The limitations are; (1) background color is black and
character color is white, (2) it is restricted to 3 letters with 1 to
4 numbers, and (2) analysis is based on one-row plate
number.
XI. IMAGE ACQUISITION & PLATE NUMBER RECOGNITION
Following flow chart shows the general step of algorithm
used for plate number recognition in this project. This
algorithm is chosen because it is suitable for number plates.
Fig 6. License plate recognition flowchart
1. RGB to binary conversion
a. Firstly, the image must be converted to grayscale
because the CCD camera captures an RGB image.
Also, the conversion is done to increase the processing
speed as grayscale image use less memory.
b. The image is resized to 800x600 pixels.
c. Then, the grayscale image is converted into binary
image which is consists of 1 and 0 digitally for each
pixel of image.
2. Image Filtration
a. Each connected component will be assigned to an
integer number from 1 to the total number of adjacent
components.
b. After the connected components have been labeled,
their width and height are been calculated. The
minimum and maximum width and height of the
labeling object (in pixels) are;
a. Heightmin = 3
b. Widthmin = 2
c. Heightmax = 70
d. Widthmin = 80
c. Each connected component that is calculated will be
compared to the referenced width and height. As for
the height values, only the connected component that
had height less than Tmax_h and greater than Tmin_h,
will be retained. Otherwise, the object will be
removed.
d. As a result, the image after this process only
contained the number plate and several objects that had
almost same size of the number plate.
e. Then, the area of each connected objects that left are
calculated. This process also used connected
components techniques. The minimum and maximum
area of the number plate are:
a. Areamin = 8
b. Areamin=1700
f. The objects that the areas within the reference areas
International Journal of Machine Learning and Computing, Vol. 2, No. 2, April 2012
95
will remain.
3. Analysis and Dilation: Then, use dilation
(morphological) operation for the resulted image.
Dilation adds pixels to the boundaries of objects in an
image. So, this process is used to combine the close
objects, by using structure element value equals to 8
pixels in 0 degree (0°). Objective of this process is to
group the characters and numbers in number plate as a 2
groups.
4. Accurate location of License Plate:
If the object remained in the image is only 1, then the
system will assume the object is the plate. But, if the object
is more than 1, eliminate all objects that are too near with
the border.
5. Character Recognition:
To recognize plate numbers, each character must be
analyzed.
a. Binarization of the recognized license plate
b. Segmentation of each character
c. Template matching to recognize license plate
numbers
6. Performance Measure:
Component heads identify the different components of
your paper and are not topically subordinate to each other.
Examples include Acknowledgments and References and,
for these, the correct style to use is “Heading 5”. Use
“figure caption” for your Figure captions, and “table head”
for your table title. Run-in heads, such as “Abstract”, will
require you to apply a style (in this case, italic) in addition
to the style provided by the drop down menu to
differentiate the head from the text.
There are several ways to recognize the characters. The
approaches are; (1) autocorrelation, (2) Mean Square Error
(MSE) and (3) Structural Similarity Index (SSI).
7. Autocorrelation:
It quantifies the closeness between two images. This
coefficient value ranges from -1 to +1, where the value +1
indicates that the two images are highly correlated and are
very close to each other.
()()
()()
22
mn mn
mn
mn mn
r
mn mn
AB
AB
AB
A
B
−−
=
∑∑
−−
∑∑ ∑∑
(1)
where
A
= mean2(A), and
B
= mean2(B)
8. Mean Square Error:
Computes the square difference between pixels in two
different images and then taken the average over all pixels
in the image. An image is perfect reproduction of original
image will have an MSE of zero, while an image that
differs greatly from the original image will have a large
MSE.
() ()
2
,
11 ,
1MM
xy
YX xy
MSE
M
NQ
P
==
=−
∑∑
m (2)
Where M, N are the dimension of the image,
()
,
x
y
p
is a
pixel of the original image and
()
,
x
y
Qis the corresponding
pixel from the reconstructed image.
9. Structural Similarity Index:
Universal quality index models any distortion as a
combination of three different factors, namely; (a) loss of
correlation, (b) luminance distortion and (c) contrast
distortion.
[
]
1, 1SSI =− + (3)
The best value 1 is achieved if and only if the two images
are similar and -1 if the images are highly not similar.
XII. PROPOSED ALGORITHM FOR CHARACTER
RECOGNITION
Since there are two databases that we have developed for
easier processes, the recognition must be done in two ways;
each one for letters and numbers. The characters are cropped
and binarized manually. Each character will be presented by
four images. These images are taken from different condition
such as from front view, slanted to some angle (not fixed),
different weather condition either sunny or rainy day, or
character with noises. The mean value of these four images
will be counted by using MSE approach.
A. ‘Full’ Display System
An LCD display is used to show the working principle of
access of barrier gate. If the ‘FULL’ is displayed, the system
would not process the car image and barrier gate would not
operate. Plus, the data acquired from ultrasonic sensors of
each parking space is calculated for the calculating free
parking spaces.
Where M, N are the dimension of the image, is a pixel of
the original image and
()
,
x
y
Qis the corresponding pixel from
the reconstructed image.
B. Structural Similarity Index
Universal quality index models any distortion as a
combination of three different factors, namely; (a) loss of
correlation, (b) luminance distortion and (c) contrast
distortion. According to equation 3 the best value 1 is
achieved if and only if the two images are similar and -1 if the
images are highly not similar.
C. Direction System
The direction system starts when the barrier gate is opened.
The database will update the status of the empty parking lots
for updating the counter for ‘FULL’ display system. Then,
Matlab is used to calculate the nearest lot that has no vehicle
to the user. Afterwards, it will display the direction towards
the empty lots to the user using LEDs.
XIII. ELECTRONICS PAYMENT SYSTEM
As mention earlier this composite system will provide the
number plate recognition and electron parking bill. At the
entry, the vehicle will stop before the entry barrier and its
presence is detected by loop sensor. The loop sensor will
initiate the camera to capture a picture of the vehicle and the
LPR module will analyze the captured picture to recognize
the number.
The captured picture together with the recognized number
International Journal of Machine Learning and Computing, Vol. 2, No. 2, April 2012
96
and entry record (entry date & time) will be stored for parking
fee calculation later. Once this is completed, the entry barrier
will open to allow the vehicle to enter and park. Another loop
sensor after the barrier will close the barrier. The entry station
is used to interface with the loop sensors and automatic
barriers.
When the leaving vehicle reach the exit booth, it will stop
before the barrier and its presence is detected by a loop sensor.
This sensor will initiate a picture to be taken and the license
plate to be read by the LPR module. The LPR module will
match the recognized vehicle number with its own file for the
entry time for this particular vehicle. Once the exit and entry
record are matched, the system will calculate and display the
parking fee is due. When the transaction is complete, the exit
barrier will open and the vehicle will leave. For barrier
control there is another barrier another loop sensor is
implemented.
XIV. RESULT
As a result, auto parking system is able to recognize plate
number, display free parking spaces and guidance parking
system. This study output is a Matlab GUI which is an
interface for users and drivers. Number plate recognition
results are shown in Table I, II and III as well as billing
window is shown in Fig. 7.
To examine the performance there are 3 approaches have
been discussed earlier, the performance analysis has been
done to identify most suitable approach for characters
recognition algorithm. The Table.I below shows the analysis.
TABLE I: PERFORMANCE ANALYSIS FOR MSE, AUTOCORRELATION AND
SSI APPROACHES
Approac
h
Total
Character
s
Recognize
d
Characters
Not
recognize
d
Character
s
Accurac
y
False
Positiv
e Rate
MSE 339 324 15 95.6% 4.4%
Autocorr 339 317 22 93.5% 6.5%
SSI 339 308 31 90.9% 9.1%
As MSE approach shows the highest accuracy, it is chosen
as the approach for characters recognition.
TABLE II: LPR DETECTION WHEN CAMERA LOCATION IN THE MIDDLE OF
THE LANE
No of detection Correct detection Error detection Success %
95 90 5 94.7
95 89 6 93.7
95 84 11 88.4
95 88 7 92.6
TABLE III: LPR DETECTION WHEN CAMERA LOCATION IN THE ONE SIDE OF
THE LANE
No of detection Correct detection Error detection Success %
80 71 9 88.8
80 70 10 87.5
80 72 8 90.0
80 71 9 88.8
Fig. 7. Matlab GUI based car parking electronic billing window
XV. CONCLUSION
In this paper the development of an automatic parking
system with license plate recognition, parking lots status and
guidance parking system and electronic billing system is
successfully implemented. The performance of the
developed of algorithms for License Plate Localization and
License Plate.
Recognition is acceptable range. The developed
algorithms accurately localize and recognize in different
location of the license plate. Electronic billing system
performance is also acceptable and recommended for
commercial use.
International Journal of Machine Learning and Computing, Vol. 2, No. 2, April 2012
97
ACKNOWLEDGMENT
Authors like to thank to International Islamic University
for arranging internal Fund for this project.
REFERENCES
[1] SherrZheng Wang, HsMian Lee, “Detection and Recognition of
License Plate Characters with Different Appearences”, in proc. Conf.
Intelligent Transportation Systems, vol. 2, pp. 979-984, 2003.
[2] Younghyun Lee1, Taeyup Song, Bonhwa Ku1, Seoungseon Jeon,
David K. Han , Hanseok Ko “ License Plate Detection using Local
Structure Patterns” in proc. Seventh IEEE International Conference on
Advanced Video and Signal Based Surveillance pp 574-579, 2010.
[3] Danian Zheng, Yannan Zhao, Jiaxin Wang, “An efficient method of
license plate location, Pattern Recognition Letters,” ISSN 0167-8655,
10.1016/j.patrec.2005.04.014. Vol. 26/15, pp 2431-2438, Nov. 2005,
[4] Preemon Rattanathammawat, Thanarat H. Chalidabhongse, “A Car
Plate Detector using Edge Information”, in proc. International
Symposium on Communications and Information Technologies, ISCIT
'06., pp.1039-1043, October 18-September 20, 2006.
[5] Xifan Shi, Weizhong Zhao, and Yonghang Shen, “Automatic License
Plate Recognition System Based on Color ImageProcessing”, Lecture
Notes on Computer Science, Springer-Verlag, Vol. 3483, pp. 307-314,
2005.
[6] E.R. Lee, P.K. Kim, H.J. Kim, “Automatic recognition of a car license
plate using color image processing”, in proc. IEEE International
Conference on Image Processing, pp. 301–305, Austin, Texas, 1994.
[7] X. He, H. Zhang, W. Jia, Q. Wu, T. Hintz, ‘Combining Global and
Local Features for Detection of License Plates in a Video’, in proc. of
Image and Vision Computing New Zealand 2007, pp. 288–293,
Hamilton, New Zealand, December 2007
[8] Otsu, N., A threshold selection method from gray-level histogram, in
journal on IEEE Transactions on Systems Man and Cybernetics,
SMC-9, pp.62-66. 1979.
[9] Qadri, M. T. and Asif, M. Automatic Number Plate Recognition
System For Vehicle Identification Using Optical Character
Recognition; 2009 in proc. International Conference on Education
Technology and Computer. IEEE Xplore; pp.335-338. 2009.
[10] Saleh, Marwan D.; Mellah, H.; Mueen, Ahmed; Salih, N. D.; , "An
efficient method for vehicle license plate xtraction," International
Symposium on Information Technology, 2008. ITSim 2008, vol.2, no.,
pp.1-5, 26-28 Aug. 2008.
[11] B. Fröba, and A. Ernst, “Face Detection with the Modified Census
Transform”, in proc. Sixth International Conference on Automatic
Face and Gesture Recognition (FGR'04), pp. 91-96, Seoul, Korea,
May 17-19, 2004.
[12] Babu, C.N.K.; Nallaperumal, K., "A license plate localization using
morphology and recognition,"
India Conference, 2008. INDICON
2008. Annual IEEE, vol.1, no., pp.34-39, 11-13 Dec. 2008.
[13] Wei-Hsun Lee, Shian-Shyong Tseng, Ching-Hung Wang, "Design and
implementation of electronic toll collection system based on vehicle
positioning system techniques" in Journal on Computer
Communications, Elsevier Science Publishers B. V. Amsterdam, The
Netherlands, Vol. 31/12, pp 2925–2933 July 2008.
[14] Faradji, F.; Rezaie, A.H.; Ziaratban, M.;, "A Morphological-Based
License Plate Location", in proc. IEEE International Conference on
ImageProcessing, 2007, Vol. 1, pp:1 - 57 - 1-60, 2007.
[15] Christos-Nikolaos E. Anagnostopoulos, Ioannis E. Anagnostopoulos
Ioannis D. Psoroulas, Vassili Loumos, Eleftherios Kayafas, “License
Plate Recognition From Still Images and Video Sequences: A Survey”
in journal on IEEE transactions on intelligent transportation systems,
vol. 9, no. 3, 377-383 September 2008.
Muhammad Mahbubur Rashid (M’07) was born in
Naogaon, Bangladesh. He received the B.Sc. (Eng.)
degree in electrical and electronic engineering from
Bangladesh University of Engineering and
Technology, Dhaka, Bangladesh, in 1992, and the
M.Sc. and Ph.D. degrees in electrical engineering
from the University of Malaya, Kuala Lumpur,
Malaysia, in 2003 and 2007, respectively. From 1994
to 2000, he was a Sub divisional Engineer
(instrumentation and control) with the Bangladesh Power Development
Board. Since 2007, he has been an Assistant Professor in the Department of
Mechatronics Engineering, International Islamic University Malaysia, Kuala
Lumpur. He has published more than 65 papers in journals and conference
proceedings. His research interests include advanced control system and
simulation and nonlinear modeling, process control and industrial
automation, instrumentation, neural networks, artificial intelligence, power
electronics, and renewable energy.
Abiodun Musa Aibinu is an Assistant Professor at the
Department of Mechatronics Engineering, Faculty of
Engineering, International Islamic University
Malaysia. His research interests are: Biomedical
Signal Processing; Digital Image Processing;
Instrumentation and Measurements; Wireless and
Telecommunication system and Digital System
Design
MD. Ataur Rahman, PhD, is an Associate Professor
in the Department of Mechanical Engineering,
Faculty of Engineering, International Islamic
University Malaysia. His research interests are hybrid
engine, intelligent power train for hybrid and
electrical vehicle, intelligent steering system and
traction control system, electromagnetic actuated
CVT and intelligent air-cushion vehicle for swamp
and peat terrain. He has worked in The University of
Tokyo, Japan, as a Visiting Fellow on the development of integrated
instrumentation systems for Autonomous Vehicles. He has published 100
Journal articles including 40 ISI listed journal from his research work.
Nur Farahana Binti Mohd Suhaimi received the B.Sc. (Eng.) degree from
Mechatronics Engineering, Faculty of Engineering, International Islamic
University Malaysia, in 2010. Currently she is pursuing her M.Sc
Engineering.
Ainul Farhana Binti Mohd Yunus received the B.Sc. (Eng.) degree from
Mechatronics Engineering, Faculty of Engineering, International Islamic
University Malaysia, in 2010. Currently she is pursuing her M.Sc
Engineering.
International Journal of Machine Learning and Computing, Vol. 2, No. 2, April 2012
98