Sei sulla pagina 1di 5

International Journal of Advanced Technology & Engineering Research (IJATER)

REAL TIME LICENSE PLATE RECOGNITION SYSTEM


Lekhana G.C, M.Tech;
Siddganga Institute of technology,Tumkur.
R.Srikantaswamy, Professor.
Siddganga Institute of technology,Tumkur.

Abstract

directed to different lanes for a better congestion control in busy


urban communications during the rush hours.

License plate recognition uses image processing and character


recognition approaches to identify vehicles by automatically
reading their license plates. In this work an intelligent real time
On Line License Plate Recognition System which caters to the
need for intelligent traffic management system, in order to cope
up with constantly increasing traffic on todays roads has been
developed. License plate recognition (LPR) algorithms in images or videos are generally composed of four processing steps,
namely, Image acquisition, License plate detection and extraction, Character segmentation, is done through fusion of spectral
analysis and connected component analysis, and Character recognition is done using support vector machines.

Anagnostopoulos [2] proposed an algorithm for vehicle license plate identification on the basis of a novel adaptive image
segmentation technique and connected component analysis in
conjunction with a character recognition neural network. ShyangLih Chang et al [2]. formulated a license plate locating module
by fuzzy disciplines while the license number identification
module was conceptualized in terms of neural subjects. . Clemens Arth and Florian Limberger [3] implemented LPR on an
embedded DSP platform that processes a video stream in realtime. Cemil Oz and Fikret Ercal [4] designed a computer vision
system to recognize license plates of vehicles in real-time environments. Lloyd Alan Fletcher and Rangachar Kasturi [5] developed an algorithm for automated text string separation which is
relatively independent of changes in text font style and size, and
of string orientation.

Keywords: Connected component, Spectral analysis , segmentation, recognition.

I. Introduction
License plate recognition systems have received a lot of attention from the research community. With the rapid growth in the
number of vehicles, there is a need to improve the existing systems for identification of vehicles. A fully automated system is in
demand in order to reduce the dependency on labour.
License Plate Recognition is a combination of image
processing, character segmentation and recognition technologies
used to identify vehicles by their license plates. Since only the
license plate information is used for identification, this technology requires no additional hardware to be installed on vehicles.
LPR technology is constantly gaining popularity, especially in
security and traffic control systems. License Plate Recognition
Systems are utilized frequently for access control in buildings
and parking areas, law enforcement, stolen car detection, traffic
control, automatic toll collection and marketing research.
LPR system can be used to calculate duration of parking.
When a vehicle enters through a gate, number plate is automatically recognized and stored in database. When a vehicle later
exits the parking area through an exit gate, number plate is recognized again and paired with the first-one stored in the database. The difference in time is used to calculate the parking fee.
Automatic number plate recognition systems installed on country
borders automatically detect and monitor border crossings. Each
vehicle can be registered in a central database and compared to a
black list of stolen vehicles. In traffic control, vehicles can be
ISSN No: 2250-3536

The focus in this work is to experiment and develop an efficient algorithm for Real Time License Plate Recognition for vehicle identification. Two main image segmentation stages are
identified in this application. First, it is necessary to capture an
image of a vehicle from video next to locate and extract the license plate region from a larger scene image. Second, having a
license plate region to work with, the alphanumeric characters in
the plate need to be extracted from the background so as to deliver them for recognition stage.

II.License plate Extraction


The first processing stage in a License Plate Recognition system is the detection and extraction of the license plate area from
a larger image, Several approaches have been proposed to solve
this problem, each having different computational expense and
success rates. The plate segmentation is one of the most important processes in the automatic license plate recognition, because
all further steps rely on it.
The two efficient methods used in the work, namely, Spectral
Analysis Approach, Connected Component Analysis are discussed in detail. The first approach makes use of the Fourier
transform to detect the inherent spatial frequency of the characters in a license plate, whereas the second method uses image
labeling based on pixel connectivity to find the boundaries of the
objects in the image.

Volume 2, Issue 4, July 2012

87

International Journal of Advanced Technology & Engineering Research (IJATER)

A. Spectral Analysis Approach


Characters in the license plate form a major license plate feature. Spectral analysis has been commonly used in segmentation
of text regions. Since the Fourier transform is complex, it is convenient to use its magnitude, also called the Fourier spectrum,
given by equation (1). Also commonly used is the power spectrum or power spectral density defined as the square of the Fourier spectrum, given by equation (2).

(a)

(b)

Figure 1: (a)Input image, (b) row wise periodogram profile.

When an image of the vehicle with M rows and N columns is


considered, each row is analyzed one at a time to see if it crosses
the license plate characters.
A periodogram which is an estimate of the power spectral density
for a series of finite length, is calculated on the intensity levels of
each row. The periodogram gives a distribution of energy per
frequency band.

The periodogram estimate is computed using FFT whose complexity is N log 2 N , which a more efficient implementation of
one dimensional DFT. is To obtain the periodogram of each row,
we first compute square of FFT's magnitude, sampling at every
pixel (sampling rate is equal to N). In order to obtain one sided
spectrum (i.e., neglecting negative frequencies), only [1, N/2+1]
components are kept. Finally, the remaining components are
multiplied by a factor of two to preserve the total energy, and
normalized by N.

In order to enhance the profile peaks, we first subtract the profile mean from the profile points. Any negative values are set to
zero, as they represent responses too low to consider and may
introduce noise in further processing. An empirical threshold of
40% of the maximum was found to work best to distinguish the
license plate candidate rows. Finally, the range of rows with values greater than the threshold is extracted from the image. The
rows containing license plate characters can be extracted with
maximum efficiency if the frequency band considered is [N/12
N/6]. A threshold of 40 % of the maximum of the averaged periodogram is set. All the rows having the value of averaged periodogram above the threshold are considered and are as shown in
Figure (2).

Figure 2: Row segmentation of an image containing license plate.

Few problematic instances usually involve other text elements


on the vehicle, manufacturers logo, presence of headlights that
detract attention from the license plate character features. Hence
an efficient method that uses connected component analysis approach along with spectral analysis approach is detailed in the
next section.

B.Connected Component Analysis Approach


In order to maximize the probability of detecting the rows in
the image containing license plate, the average of the periodogram values over a frequency range [u1, u2] is computed, where
the license plate character energy is expected to dominate. This
average represents the strength of the frequency components in
the signal (image row) sharing such periodicities. By plotting
these average strengths in a row-wise profile, the row locations
containing license plate characters produces distinctive peaks as
shown in Figure(1).

Connected components scans an image and groups its pixels


into components based on pixel connectivity, i.e. all pixels in a
connected component share similar pixel intensity values and are
in some way connected with each other. This method requires the
input image to be converted into binary for processing. The
boundary of each object is checked to see if it occupies expected
area. This method makes use of the property that all license
plates are rectangular and the characters on them share the same
background.
This method may result in one or more regions which are most
likely to resemble license plate depending on the predefined relative dimensional constraints. This might also lead to the selection
of other candidates which may not contain characters but have

ISSN No: 2250-3536

Volume 2, Issue 4, July 2012

88

International Journal of Advanced Technology & Engineering Research (IJATER)

the dimension of a license plate. Hence spectral analysis is then


applied on these license plate candidates to extract the region
containing the license plate. If none of the regions response
matches that of a standard license plates response, then only
spectral analysis approach is used. Hence, this method improves
the efficiency in cases where spectral analysis approach might
over shoot in selection of exact license plate region.

image with an object in white colour on a black background at


pixel level. In this analysis, all the objects in the image are
labelled using the concept of adjacency of the pixels. In this case
there are two objects. The coordinates of the pixels
corresponding to the labelled region are analyzed to find the
limits which separate an object from other objects if present in an
image. Figure 4 shows the segmented character of Figure 3.
SEGMENTED CHARACTERS

Input iname

Figure 4: segmented character of Figure (3).

(a)

IV.Character Recognition

(b)

License plate candidates

25

20

15

10

-5
0
50
100

The license plate of a vehicle contains alphanumeric characters which are extracted using different methods discussed in the
previous sections. Recognition of such segmented characters is
the last operation in the system where the input from the feature
extraction operation is being translated and understood by the
system.

150
200
250
300
350
400

(c)

LP

(d)

A. Binary Classification

(e)
Figure 3: (a)input car image(b) Binary image of car (c) License plate
candidates, (b) Selection of the license plate using spectral analysis
approach, (e) Segmented license plate

The license plate appears as a white rectangle, but there are


several other white regions as well. While selecting the regions
care is taken to see that regions with width and height close to
the actual image size are avoided.
The sequence of steps followed using proposed method i.e., fusion of spectral analysis and connected component analysis approach as applied to a test image is shown in Figure(3)

SVMs belong to the class of maximum margin classifiers. A


support vector machine constructs a hyperplane or a set of
hyperplanes in a high or infinite dimensional space, which can be
used for classification, regression or other tasks as shown in Figure (5). They perform pattern recognition between two classes by
finding a decision surface that has maximum distance to the closest points in the training set which are termed support vectors.
Intuitively, a good separation is achieved by the hyperplane that
has the largest distance to the nearest training datapoints of any
class (so-called functional margin), since in general the larger the
margin the lower the generalization error of the classifier [18]. In
the Figure, H2 forms the optimal separating hyperplane (OSH)
since it has maximum distance to the closest datapoints.

III.Character Segmentation
The decomposition of license plate images into spatially coherent regions in which the characters are distinguished from the
background is an image segmentation problem.
Connected component analysis approach: Approach for character
extraction is the connected component analysis. This was also
used in the fusion technique of extracting license plate. It is
explained as applied to character extraction here. The same
explanation holds good for extracting license plate candidates
also. The input image must be binarized such that the objects of
interest in it are mapped to white on a black background. An
ISSN No: 2250-3536

Figure 5: Binary classification in SVM

Consider a training set of points xi Rn , i= 1, 2,........N


where each point xi belongs to one of the two classes identified
by the label yi ( x ). (y ) {-1,1}. Assuming linearly separable

Volume 2, Issue 4, July 2012

89

International Journal of Advanced Technology & Engineering Research (IJATER)

data, the goal of maximum margin classification is to separate


the two classes by a hyperplane such that the distance to the support vectors is maximized. This hyperplane is called the optimal
separating hyperplane (OSH). The OSH has the form:

(5)
The coefficients i and b in Eq (5) are the solutions of a quadratic programming problem. Classification of a new data point
xi is performed by computing the sign of the right side of Eq (5).

(a)

(b)

(c)

(d)

Figure 6: (a) (c) Test character; (b) (d) Character recognition using
SVM

In the following, Eq (6) is used to perform multi-class classification. The sign of d is the classification result for x, and |d| is
the distance from x to the hyperplane. Intuitively, the farther
away a point is from the decision surface, i.e. the larger |d|, the
more reliable the classification result
Figure 7:shows successful character recognition of Figure3(a)

Experimental Result
(6)
The entire construction can be extended to the case of nonlinear separating surfaces. Each point x in the input space is
mapped to a point z = ( x) of a higher dimensional space,
called the feature space, where the data are separated by a hyperplane. The key property in this construction is that the mapping
(.) is subject to the condition that the dot product of two points
in the feature space ( x). ( y) can be rewritten as a kernel function K(x,y).
The decision surface has the equation
(7)

again, the coefficients


and b are the solutions of a quadratic
programming problem. Note that
does not depend on the
dimensionality of the feature space. An important family of kernel functions is the polynomial kernel:
(8)
where d is the degree of the polynomial. In this case, the components of the mapping
are all the possible monomials of
input components up to the degree d. In Figure 6 shows the success full recognition of characters using support vector machines

Figure 8:shows the success full recognition of a moving vehicle.

License plate detection is accomplished using fusion of Spectral Analysis and Connected Component Analysis. The characters are extracted from the license plate using Connected Component Analysis and recognition is accomplished using SVM.
The following Figure 8 shows the success full recognition of a
moving vehicle.

V.Conclusion
An On Line License Plate Recognition System has been developed and tested. Spectral Analysis approach and Connected
Component Analysis approach are generally used for license
plate extraction. It is found that when these techniques are used
individually, they failed. However, fusion of both Spectral Analysis and Connected Component Analysis gives better results. For

ISSN No: 2250-3536

Volume 2, Issue 4, July 2012

90

International Journal of Advanced Technology & Engineering Research (IJATER)

character segmentation , Connected Component Analysis gives a


better performance over projections method for segmentation of
characters from license plate.
SVM feature extraction techniques for character recognition is
made in this work. It is found that SVM classifier gives better
recognition accuracy compare to PCA LDA feature extraction
method.

[12]

David C. Lay, Linear algebra and its applications,


Third edition, Pearson Education, 2006

[13]

R. Chellappa, C.L. Wilson, and S. Sirohey, Human and


machine recognition of faces: A survey, Proceedings of
IEEE, Vol. 83, pp. 705740, 1995

[14]

Mathew.Turk and A. Pentland, Eigenfaces for Recognition, Cognitive Neuroscience, Vol. 3, pp.71- 86,
1997.

[15]

E. Tzafestas, A. Nikolaidou and S. Tzafestas, Performance Evaluation and Dynamic Node Generation Criteria for Principal Component Analysis, Neural Network,
Mathematics and Computer in Simulation, Vol. 51,
pp.145- 156, 300, 2004

[16]

Roy, A.; Ghoshal, D.P. Number Plate Recognition for


use in different countries using an improved segmentation IEEE Conference Publications 2011

[17]

Yifu Wang; Peng Xu Vehicle License Plate tilt


correction based on the Weighted Least Square
Method IEEE Conference Publications 2010.

VI.References
[1]

Kasaei, S.H.M.; Kasaei, S.M.M., Extraction and Recognition of the Vehicle License Plate for Passing under
Outside Environment IEEE Conference Publications
2011.

[2]

C.N. Anagnostopoulos, I. Anagnostopoulos, V. Loumos, and E. Kayafas, A license plate recognition algorithm for Intelligent Transportation System applications.

[3]

Shyang-Lih Chang, Li-Shien Chen, Yun-Chung Chung,


and Sei-Wan Chen , Automatic License Plate Recognition, IEEE transactions on intelligent transportation systems, Vol. 5, March 2004

[4]

Clemens Arth, Florian Limberger, Real-Time License


Plate Recognition on an embedded DSP-Platform

[5]

Cemil Oz, Fikret Ercal , A Practical License Plate


Recognition System for Real-Time Environments

[6]

Lloyd Alan Fletcher, Rangachar Kasturi, A Robust


Algorithm for Text String Separation from Mixed
Text/Graphics Images, IEEE Transactions on Pattern
Analysis and Machine Intelligence, Vol. 10, No. 6, November 1988

[7]

Varsha Kamat, Subramaniam Ganesan, An Efficient


Implementation of the Hough Transform for Detecting
Vehicle License Plates Using DSPs,
and pattern recognition, neural networks and image
processing.

[8]

Yasuharu Yanamura, Masahiro Goto, Daisuke Nishiyama, Extraction and Tracking of the License Plate
Using Hough Transform and Voted Block Matching

[10]

Beatriz Daz Acosta, Experiments in Image Segmentation for Automatic Us License Plate Recognition

[11]

Rafael C.Gonzalez and Richard E.Woods, Digital image processing using MATLAB, Second Edition, Pearson Education, 2006

ISSN No: 2250-3536

Biographies
LEKHANA G.C Persuing M.Tech in Siddaganga Institute of
technology,Tumkur,Karnataka, received the BE. degree in Telacommunication Engineering from the Visvesvaraya Technological University Belgaum, Karnataka, in 2010, and research areas
include Imaga processing,pattern recognition, Image enhancement. Lekhana G.C may be reached at lekna.tc@gmail.com.
R. Srikantaswamy received his M.Tech degree in Industrial
Electronics in 1995 and Ph.D. degree in Electronics in 2006 from
University of Mysore, India. He is working as a Professor in the
Department of Electronics and Communication, Siddaganga Institute of Technology, Tumkur, India. His research interests include computer vision and pattern recognition, neural networks
and image processing. Dr.Srikantaswamy may be reached at
rsswamy@rediffmail.com.

Volume 2, Issue 4, July 2012

91

Potrebbero piacerti anche