Вы находитесь на странице: 1из 8

AbstractThe ASS (Automobile Surveillance System) is directed to provide with a security which is hard to be counterfeited.

ASS uses
image processing technology to authenticate vehicles by their license plate and organizations logo printed on sticker. The system is intended to
help in the recognition of number plates of vehicles and sticker matching. This system is based on image processing techniques. The proposed
system helps in the functions like detection of the number plates of the vehicles along with the sticker (attached to inner surface of windscreen),
processing both of them individually and using processed data for further processes like storing, validating (Database matching) and permitting
vehicle to pass or rebuff vehicle. It also helps in the reduction of space consumption, by conducting the graphic images of the vehicles which can
be further stored in the database in text format reducing size of data to be stored. The system is implemented in Matlab involving Ms-Access as
backend.

Index TermsDigital image processing, License plate recognition, Logo matching, Optical character Recognition (OCR), Template matching,
Validation.



1 INTRODUCTION
URING the earlier part of the century most of the
security for entrance was done manually which
involve a lot of resources (Physical or HR) and time
for matching/checking of the Vehicles data. Evolution of
new technologies smoothened the way for automatic data
collection, storage, and retrieval in different areas of
concern. It has also helped in automation together with the
validation of the security of vehicles. Manual security of
automobile is still in trend.
Many institutes and organizations in Pakistan are still
following manual system for detection of stickers/tag at
entrance (for vehicles). This manual system has many
drawbacks, e.g. anyone can create a carbon copy of the
tag/sticker and affix to the inner surface of the windshield
of his car and enter in to the organization.
The anticipated system is able to analyze two images of
car captured from the digital camera, one is used to trace
the registration plate; recognize registration number of the
car and authenticates it, while the other one is required to
validate organizations logo. The research approach
includes locating license plate and organizations logo in
the digital image and the segmentation of individual
characters in the license plate number.
The Automobile/Vehicle Surveillance System using
license plate numbers along with organizations tag/
sticker provides with security at national as well as
domestic level. It is complicated as compared to manually
record license plate numbers at site. As mentioned in [1],
Manual reading of license plate numbers from videotape is
pretty much time consuming.
The paper presented provides with the algorithm which
increases in efficiency level in the use of time and labor, and
possible higher accuracy rate (of security); which may be
used as a foundation for intelligent infrastructure like
traffic/speed control, harbor cargo control, e-payment etc.
2 LITERATURE REVIEW
In general, the automobile surveillance system or vehicle
identification system follows three basic steps: localization
of license plate region, segmentation of license plate
characters and recognition of every single character. This
section covers the contributions of different researchers
following this three step frame work.
2.1 License Plate Localization
For the extraction of license plate region, the technique
based upon novel adaptive image segmentation technique
is used for the segmentation (SCWs) along with connected
component analysis [2]. Comparison of RGB color
processing of every pixel with a predefined range is
considered after applying Homography (mathematical
science of geometry) to remove distortion [3]. The primary
focus of study in [4] is smearing and edge detection
algorithms. Morphological operations (opening and
closing) using dilation and erosion are implemented [5-6].
The technique used in [7] follows x-axis and y-axis
projection. For identification of varying license plate colors
histogram analysis technique is carried out, where white
characters on black background are dealt by segmentation
algorithm, if condition fulfils, the image is forwarded for
further processing otherwise the image is first reversed and
then forwarded [8]. Canny edge detection method along
with Extended Hough transform (EHT) is used [9].
Automobile Surveillance System
Ammara Zulfiqar, Asim Munir, Memoona Khanam, Malik Sikandar Hayat Khiyal
D

Ammara Zulfiqar is an under graduate student of Department of
Software Engineering, Fatima Jinnah Women University, The Mall,
Rawalpindi, Pakistan.
Asim Munir is Assistant Professor at Computer Science
Department at International Islamic University Islamabad,
Pakistan.
Memoona Khanam is Lecturer at the Department of Computer
Sciences, Fatima Jinnah Women University, The Mall, Rawalpindi,
Pakistan.
Dr. Malik Sikandar Hayat Khiyal is Professor and Chairman of D
Jinnah Women University, The Mall, Rawalpindi, Pakistan.

JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 138
2.2 Character Segmentation
The license plate region localized in prior section follows
the segmentation of characters present in the localized
plate. The segmentation of license plate characters done in
[2] is based upon SCWs technique along with bounding box
technique. Projection algorithms are implemented [3,8]. The
key focus of [4] is smearing algorithm along with some
other morphological operations. Morphological operations
are applied [5]. Sunhee Kim [6] proposed Gibou-Fedkiw
algorithm. Projection algorithms using Hough Transform is
applied [7]. Discrete wave transform using multi resolution
analysis concept is proposed in [9] for segmentation of
characters.
2.3 Character Recognition
The license plate character segmentation done in
preceding section is followed by the recognition of those
characters. The algorithm implemented in [2,9] is
Probabilistic Neural Network consisting of three layers
Input, Hidden and Output. The recognition rate reported in
[2] exceeds 90%.The input vector with 180 nodes, middle
layer consists of 108 nodes and the output layer is
composed up of 36 nodes. Optical Character Recognition
using correlation technique is implemented in [3,6] with an
impressive performance ratio 92.57% [3]. Statistical based
template matching is used for recognition purpose of
known images; whereas cross correlation for matching
correlation coefficient of known image with the unknown
[4]. The main focus of [8] is two Back Propagation Neural
Networks; for the purpose of recognition of license plate
data; with a remarkable recognition rate 96.84%.
3 FRAMEWORK OVERVIEW
4 TECHNIQUE
The Automobile Surveillance System as presented in fig.1,
proposed in this paper follows a series of steps, consisting
of three distinct parts. The first one deals with the
successful data entry, deletion or modification with in
database. The second includes operations for the validation
of license plate. The third part consists of verification of
sticker affix to inner surface of windshield.
4.1 Administrator Controls
The GUI of the system is composed of admin login and
processing/gate controls. Admin login requires login
password to enter in to the window. If correct password is
entered the system shifts the control to admin window to
modify database records. Visual Query Builder is the tool
used for bridging the gap between frontend (Matlab) and
database (Ms-Access) the run time changes done in Matlab
get stored in Ms-access database. The admin level pseudo
code is given in table I.
Fig. 1. Block Diagram of proposed Automobile Surveillance System

TABLE I
ADMIN LEVEL PSEUDO CODE
Input: Text
Admin Controls
1 Enter Password
2 If correct
{
Login to Admin Window
}
Else
{
Display Error Box: Wrong Password, Retry.
}
3 Click on the modify database button
4 Select option from Query builder
If Select
{
Select table and then columns to be displayed.
Assign a MATLAB workspace variable against
which the values are executed
}
Else If Insert
{
Insert values in sequence as there are names in
the database table against any variable.
Execute that variable in Query builder.
}
5 If matched
{
Display Message Box : Data in variable
<variable name> successfully written to
table <table name>
}
Else
{ Display Error Box
}
Output: Message Box
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 139
4.2 License Plate Processing
The License plate processing is further divided in to three
phases. The first one deals with the localization of license
plate. The second one includes successful segmentation of
license plate characters where as the third phase consists of
recognition of segmented characters. The license plate
processing pseudo code is given in table II.

a) License Plate Localization: The localization of
license plate follows some sequence of steps. First image is
converted to grey scale based on equation (1):

Intensity 0f uiay Sclae Image =
R + u +B
S
(1)
The grey scale image is transformed to binary image
using equation (2):

g(x, y) = _
(x, y) > I
(x, y) I
(2)

The binary image is used for connected component
labeling. The 8- neighbors are found using following
equation:
N
8
(p) = N
4
_
(x +1, y +1), (x + 1, y - 1),
(x - 1, y + 1), (x -1, y -1)
_ (S)
By using area properties from regional descriptors and
SCW technique the license plate region is identified. The
aspect ratio and area is specified based upon equation (4)
and (5) (see fig. 2,3 for output).
Aspcct Rotio =
X
B
-X
A
((
B
-
A
) 2)
(4)
Arco = Icngtb wiJtb (S)
The license plate is cropped using the maximum and
minimum values of x and y axis, whose result is shown in
fig 4.







b) Character Segmentation: The segmentation of license
plate follows same sequence of steps as stated in the prior
section for the extraction of license plate. Only the
difference occurs in the specification of aspect ratio and
area. The output is shown in fig. 5, 6.






Fig. 3. Grey Scale Image with
Bounding Box


Fig. 2. Original Image


Fig. 4. Extracted License Plate


Fig. 5. License Plate with
Bounding Box
TABLE II
LICENSE PLATE PROCESSING PSEUDO CODE
Input: Image J
License Plate Localization:
1 Binarize input image % Specified threshold value
2 Fill image regions using imfill command
3 Label the connected components
4 Calculate Area using Bounding Box
5 Apply SCW technique % Parameters X1,X2,Y1,Y2, BB
6 Obtain objects fulfilling following conditions
(Area > 5,000 and Area < 40,000) and
(Aspect Ratio > 2 and Aspect Ratio < 2.7)
7 Identify X min, Xmax, Ymin, Ymax% LP acquired
8 Write the Obtained Object

Character Segmentation:
9 Read the image
10 Complement Image
11 Label the connected components
12 Calculate Area using Bounding Box
13 Apply SCW technique % Parameters X1,X2,Y1,Y2, BB
14 Retrieve objects fulfilling following conditions
15 (Area > 100 and Area < 1,000) and
16 (Aspect Ratio > 0.35 and Aspect Ratio < 0.6)
17 Identify X min, Xmax, Ymin, Ymax % Segment Char
18 Write the Obtained Object
18 Count A; % Number of characters/digits
19 for B=1:A % A is image
{
Store in Array
Forward to OCR for Recognition }

Character Recognition:
20 Open Text File % Write mode
21 Resize characters to 42*24 pixels % Template size
22 Compute Correlation
23 Close Text file

Database Matching:
24 Read Text File % Recognized characters
25 Fetch Database Records % Stored Characters in DB
26 Search for Match
27 if matched
{
Display Message Box : License plate matched
Successfully }
Else
{
Display Message Box : License plate not matched }
Output: Message Box

JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 140

c) Character Recognition: The recognition of
segmented characters follows OCR algorithm using
template matching technique. (see fig. 7 for output)











d) Database Matching: The recognized License plate
characters stored previously in the form of text file is
further processed for the matching purpose. The characters
are matched with the ones stored in Ms-Access database, as
shown in fig (8). (see figure (9) for results)

4.3 Sticker Processing
The sticker recognition also follows some of the
preprocessing steps as all the previous phases have done
but in this phase the preprocessing is carried out till the
image is converted to gray scale.
The Simple noncircular correlation method for
exhaustive sum square difference matching, template
matching technique is used for the purpose of sticker
matching. (see fig. 10-13 for output)




5 EXPERIMENTAL RESULTS
The system is completely tested by developing the test
cases.
The system is tested using a selected set of 65 arbitrary
data samples collected from different area, also composed
of samples from different provinces of Pakistan for license
plate recognition out of which result of 7 samples are shown
in table V (at the end) whereas 30 samples for the purpose of
sticker matching out of which result of 6 samples are shown
in table VI (at the end) and the overall system accuracy is
shown in table IV. Stickers templates were considered
from two different organizations. The tested samples
results are match with are analysis or the assumptions
assumed to check the accuracy of system, as shown in fig.
14, 15.


Fig. 6. Segmented License Plate Characters


Fig. 7. Recognized License
Plate Characters


Fig. 8. Database records used for matching


Fig. 9. Message Box Authenticating
License Plate
TABLE III
STICKER PROCESSING PSEUDO CODE
Input: Image
Sticker Matching:
1 Convert input image to gray scale %Specified threshold
Apply SSDXCORR technique
(Simple noncircular correlation method for
Exhaustive sum square difference matching)
3 Rotate the matrix to 90 degrees
4 Convolve the input image with the conjugate of stored
template
5 Get the real part of the convolution result
6 Determine Equivalent subscript values from the
result of step 5.
7 Draw line around the matched area if any
8 if matched
{
Display Message Box : Sticker matched successfully
}
Else
{ Display Message Box : Sticker not matched
}
Output: Message Box


Fig. 10. Frame image
(Input)


Fig. 11. Template
Image


Fig. 12. Template
matching


Fig. 13. Message Box Authenticating
Organizations Sticker
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 141
5.1 Accuracy Results of License Plate Localization
The license plate localization as observed by the system
has a remarkable accuracy rate of 100%.
5.2 Accuracy Results of Character Segmentation
The accuracy rate of number plate localization, as
monitored by the system has impressive performance ratio
of 100%.
5.3 Accuracy Results of Character Recognition
The recognition of characters as examined by the system
has a success ratio of 96%.
5.4 Accuracy Results of Sticker Recognition
The accuracy rate as monitored by the system resulted in
92%.
Overall system performance observed by the system has
an impressive ratio of 97%.
6 LIMITATIONS
The proposed system has some restrictions. The system is
affected by appearance and illumination conditions.
Misclassification is faced in recognition of some of the
characters due to their likelihood.
Another limitation of the system anticipated is its
restriction to specified font and format of license plate i.e.
Time new Roman.
7 CONCLUSION
In todays world there is a great need of security. An
attractive feature of this project is that the anticipated
system provides with two level security, as it deals with not
only the recognition and validation of License Plate but also
authentication of organizations sticker.
The Application possibilities of this design are vast and
rather numerous. Ranging from Security to standard
commercial sectors, this device makes a promising stand.
The main theme of the project was gathering and
description of information and the goal was to collect
physical data, represent them symbolically, and
demonstrate processing techniques for that data. This goal
has been accomplished in this project.
A trial product has been implemented in MATLAB 7.0 on
a p4 PC under Windows XP.
The overall performance achieved remarkable success
rate i.e. 97 % just as for the individual tasks.
In general, the conclusion of this report is that a system
for Automobile Surveillance System can be constructed for
automatic traffic control system, hence reducing the labor
work and eliminating trivial tasks involved in manual work
which have been proposed in this system.
We have successfully designed and implemented the key
element of the system, the actual recognition and validation
of the license plate along with organization sticker.
With the passage of time everything requires
improvement and security is such a domain which cannot
compromise on perfection. So some enhancements need to
be considered in the projected system also.
In future, an automated system may be implemented
using RFID (Radio Frequency Identification).
This technique can further be extended by focusing
on the image of the driver and with the validation of
license plate it also validates the owner of the car
before entrance in an organization.
The system can be redesigned in future, using
Artificial Neural Network techniques for recognition
to increase the success rate of license plate characters
recognition.
REFERENCES
[1] Kelvin C.P.Wang, Advanced Planning Surveys Using Automatic
License Plate Reader, Department of Civil Engineering,
University of Arkansas, Fayetteville Nov 14, 1997.
[2] Christos Nikolaos E. Anagnostopoulos, Ioannis E.
Anagnostopoulos, Vassili Loumos, Eleftherios Kayafas A
License Plate-Recognition Algorithm for Intelligent
Transportation System Applications, IEEE Transactions on
TABLE IV
ACCURACY RESULTS
No. of
Sampl
es
Quantity
Accuracy In
Percentage
Average Elapsed Time
For Processing (ms)
65 License plate extraction 100% 38
65 Character segmentation 100%

105
65 Character recognition 96% 153
30 Sticker matching 92% 15

Fig. 14. Accuracy graph of Number Plate Recognition Made by system
for 65 Samples
0
20
40
60
80
100
120
1 9 17 25 38 50 65
A
C
C
U
R
A
C
Y

(
o
u
t

o
f

1
0
0
%
)
INPUT IMAGE SAMPLES (65)
ACCURACY GRAPH OF LICENSE PLATE RECOGNITION

Fig. 15. Accuracy graph of Sticker Matching Made by system
for 30 Samples
0
20
40
60
80
100
120
1 7 13 15 21 27 29
A
C
C
U
R
A
C
Y

(
o
u
t

o
f

1
0
0
%
)
INPUT IMAGE SAMPLES (30)
ACCURACY GRAPH OF STICKER MATCHING
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 142
Intelligent Transportation Systems, Vol. 7, No. 3, September 2006,
pp. 377392.
[3] Shishir Kumar, Shashank Agarwal & Kumar Saurabh, License
Plate Recognition System for Indian Vehicles, International
Journal of Information Technology and Knowledge Management,
July-December 2008, Volume 1, No. 2, pp. 311-325.
[4] Serkan Ozbay, and Ergun Ercelebi, Automatic Vehicle
Identification by Plate Recognition, World Academy of Science,
Engineering and Technology Sep 2005
[5] JR Parker, Pavol Federl, An Approach to License Plate
Recognition, Laboratory for Computer Vision, University of
Calgary
[6] Sunhee Kim, Seungmi Oh and Myungjoo Kang, Mathematical
Image Processing For Automatic License Plate Recognition
System, J. Ksiam Vol.14, No.1, 5766, 2010
[7] Yungang Zhang, Changshui Zhang A New Algorithm for
Character Segmentation of License Plate, Dept. of Automation,
Tsinghua University, the Institute of Information Processing
[8] Chien-Tsai Tsai and Daw-Tung Lin, Automatic License Plate
Recognition System, 18th IPPR Conference on Computer Vision,
Graphics and Image Processing (CVGIP 2005)
[9] Vijaya Laxmi, B.M Karan, Vehicle Identification Automation
Using Wavelet-Probabilistic Neural Network Combined
Approach, International Journal of Signal & Image Processing
(Vol. 1-2010/Iss.2), pp. 62-68.

Ammara Zulfiqar is the under graduate student of Department of
Software Engineering in Fatima Jinnah Women University the Mall,
Rawalpindi, Pakistan.

Asim Munir is enroll in Ph.D. He has done M.S. in Computer Science
and MSc in Computer Science. He is Assistant Professor at Computer
Science Department at International Islamic University Islamabad. He
is Gold Medalist Microsoft Official Curriculum Training-MCSE

Memoona Khanam has done MS in Computer Sciences and M.Ed.
She is Lecturer at Computer Science Department at Fatima Jinnah
Women University. She has expertise is in Aritificial Intelligence. She
has 4 years teaching experience

Dr. M. Sikandar Hayat Khiyal born at Khushab, Pakistan. He is
Chairman Dept. Computer Sciences and Software Engineering in
Fatima Jinnah Women University Pakistan. He Served in Pakistan
Atomic Energy Commission for years and involved in different research
and development program of the PAEC. He developed software of
underground flow and advanced fluid dynamic techniques. He was
also involved at teaching in Computer Training Centre, PAEC and
International Islamic University. His area of interest is Numerical
Analysis of Algorithm, Theory of Automata and Theory of Computation.
He has more than hundred research publications published in National
and International Journals and Conference proceedings. He has
supervised three PhD and more than one hundred and thirty research
projects at graduate and postgraduate level. He is member of SIAM,
ACM, Informing Science Institute, IACSIT. He is associate editor of
IJCTE and Co editor of the journals JATIT and International Journal of
Reviews in Computing. He is reviewer of the journals, IJCSIT, JIISIT,
IJCEE and CEE of Elsevier.
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 143
TABLE V
TESTING RESULTS OF LICENSE PLATE PROCESSING
(a) (b) (c) (d) (e) (f)
SR.
#
Original Image
License Plate
Localization
Character Segmentation Digit Recognition Database Matching
1

Threshold 140

Threshold 140

Threshold 140

Threshold 140

Threshold 190

Threshold 159

Threshold 144

JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617


https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 144
TABLE VI
TESTING RESULTS OF STICKER MATCHING
(a)
Sr. #
(b)
Original Sticker Input Image
(c)
Template
(d)
Sticker Recognition
(e)
Sticker Matching
1


JOURNAL OF COMPUTING, VOLUME 3, ISSUE 12, DECEMBER 2011, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 145

Вам также может понравиться