You are on page 1of 5

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882

Volume 3, Issue 5, August 2014

Number Plate Extraction Using K-means Clustering and Weighted Sum


Method
Vaishali Gupta1, Balraj Singh Sidhu2
Electronics and Communication Engineering Department
Giani Zail Singh Punjab Technical University Campus, Bathinda, Punjab, India
1, 2

ABSTRACT
In todays world automatic vehicle number plate
extraction plays a significant role. It is widely used in
surveillance systems. In this paper the fast and real time
method for number plate extraction has been proposed.
In this proposed method, firstly, original image has been
taken, and converted into grey scale image. K-means
clustering has been applied, vertical and horizontal edges
are extracted from the image, morphology operation is
used and the feature extraction is done, the correct
location of plate has been extracted using weighted sum
method (WSM) and if there is any tilt in number plate
that tilt is removed away, and extracted number plate is
displayed as output. This method has been implemented
on images with different background and distances, and
as compare to other methods improved results has been
obtained.
Keywords- edge detection, K-means clustering, number
plate extraction, plate tilt correction and weighted sum
method.

1. INTRODUCTION
Licence plate recognition system is the technology that
enables the computer systems for automatically
extraction of the licence number plate of vehicles from
digital pictures. Automatic detection and tracking is a
very challenging problem in computer vision with
various important practical applications in surveillance
systems. Manually reviewing and verifying large amount
of data is often impractical in todays world.
The licence plate recognition system works in
four
steps
Plate
Localization,
Preprocessing,
Segmentation, and Normalization. Morphological
operations have been applied to the image to identify the
exact plate location. It has a broad range of applications
such as: Tolling, Traffic control, to find out stolen
vehicles, Airport Parking and so on. [1]
Vehicle number plate recognition system is an
important part of modern intelligent traffic control
system, which can be applied to all vehicle management

systems. Licence plate recognition system consists of


computer information management technology and
digital camera technology, with advanced graphic
processing, and artificial intelligence technology, it deals
with image acquisition and processing so as to obtain
more information, to reach at a higher level of intelligent
management and accuracy. [2]
Licence plate recognition system is used in day
to day life, in order to enhance the quality of life. The
increase in number of vehicle causes some management
problems, such as: construction vehicles, company
vehicle management. A number of vehicle management
systems make use of human resources. An automatic
Licence plate recognition system is developed in order to
save labour costs and improve efficiency. [3]
Licence plate recognition system is one of the
most widely used technologies in intelligent traffic
control areas. There are large application areas forecast
in intelligent car park management, automatic toll
collection at highways and any kind of violation
detection. In recent years licence plate recognition is an
important and wide area research topic in the transport
zone. It combines computer technology, pattern
recognition, image processing and other subjects
together. [4]
Images are the most important means of
transmission of information. Analyzing images and
extracting the information from them in such a way that
the information can be used for further tasks is an
important factor of learning of machine. Clustering is the
method of grouping samples in such a way that the
samples are alike within each group. The groups are
called clusters. In K-means clustering there are always K
number of clusters, and least one item in each cluster.
Every member of a cluster should closer to its cluster
than any other cluster because closeness does not always
involve the centre of clusters. The advantages of KMeans algorithm are: It is easy to implement, simple,
quite efficient and can apply even on large data sets.[5]

www.ijsret.org

869

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 3, Issue 5, August 2014

Weighted sum method is widely used for multiobjective optimizations. It is used broadly to offer
multiple solution points by varying the weights again
and again, and also provides a single solution point that
reflects preferences apparently included in the selection
of weights of a single set. This method results a linear
approximation of the preference function. No issue how
the weights are set. The result depends on various
factors, such as the relative magnitude of the objective
functions. When weights are set, only the relative
significance of the objectives should be considered.
When ranking method is used to set weights then all the
objective functions should be changed such that they
have same ranges. This is easy to use and basic method
for multi-objective optimization. [6]
A morphology based method for licence plate
identification is also applicable in the transport
information systems, where automatic detection of
registration plates, signs, and so on is often necessary.
[7]
New edge detection and connected components
based method has been implemented on the images with
different backgrounds, various atmospheric and light
conditions and low quality images. [8]

6. If there is no tilt then go to step 8.


7. If there is any tilt in the number plate that tilt is
removed away using tilt correction algorithm.
8. After above steps the extracted number plate with tilt
correction appears as an output.
9. After applying this procedure to all images of the data
set the calculation of Percentage Efficiency of Correct
Number Plate Extraction is done.

In this paper, an algorithm for vehicle number


plate extraction has been proposed using k-means
clustering and weighted sum method and the result has
been compared with existing technique such as Edge
Detection and Connected Component Method for
Licence Plate Recognition.

2. PROPOSED WORK
In the proposed method, original image has been taken
from database and converted into grey scale image. Then
K-means clustering has been applied, vertical and
horizontal edges of the image has been extracted,
morphology operation and the feature extraction has
been done, the location of plate has been extracted,
weighted sum method has been used and if there is any
tilt in number plate that tilt has been removed away, and
extracted number plate is displayed as output. The block
diagram of the proposed method is shown in Fig. 1.
Steps for proposed algorithm are:
1. Take an input image and convert it into grey scale
image.
2. Apply k-means clustering.
3. Apply horizontal and vertical edge detection
technique.
4. Apply morphological operations.
5. Apply weighted sum method.

Figure 1: Block diagram of the proposed algorithm


2.1 Edge detection:
Edge detection is the technique used for a set of
mathematical methods used to identify points at which
the digital image brightness changes sharply or has
discontinuities. This feature is used to find plate location
in the picture. Sobel edge detection is used because it
has more positive performance as compared to other
methods due to its high speed. In this proposed method
vertical and horizontal edge detection have been
implemented. The sobel vertical mask and sobel
horizontal mask is shown in Fig. 2(a) and Fig. 2(b).

www.ijsret.org

870

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 3, Issue 5, August 2014

Figure 2(a)
Figure 2 (b)
Figure2: (a) Sobel vertical mask, (b) Sobel horizontal
mask
For edge detection, on an input image sobel horizontal
and vertical masks have been applied and absolute
magnitude of gradient is calculated.
2.2 K-means method:
Let X={xi}, i=1,2,,n is the set of n d-dimensional
points which are to be clustered into set of K clusters, C
= {ck, k = 1,2,..,K}. The purpose of K-means is to
minimize the sum of squared error over K clusters, the
squared error between points of cluster ck and k is given
in equation (1):
K
J(C) = ||xi - k||2
k=1 xick

Figure 3: Input image (sample_1.tiff)

(1)

Where: xi be the set of points, K be the number of


clusters and k be the mean of cluster ck.
The k-means method is the most widely used clustering
algorithm and is usually effective and provides good
results.

Figure 4: Image obtained after k-means clustering

2.3 Weighted sum method:


Weighted sum method (WSM) is used in decision
theory. It is the multi-criteria decision analysis used for
evaluating a number of alternatives in terms of number
of decision criteria.
n
AiWSM-score = wj aij, For i = 1, 2, 3,.., m.
J=1

(2)

For m alternatives and n decision criteria, wj denotes the


relative weight of importance; aij is the performance
value of alternative Ai; total importance of alternative Ai
is denoted as AiWSM-score as given in equation (2).
2.4 Stepwise results of the figures obtained while
implementing the algorithm have been shown below:

www.ijsret.org

Figure 5: Image obtained after edge detection

871

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 3, Issue 5, August 2014

Figure 9: Extracted number plate


Figure 6: Image obtained after morphological operation

The method used in this paper has been implemented on


150 colored images of different sizes from the data set
[9]. These images are different in point of view, different
light balance, different distances and different
backgrounds. The result of the proposed method for
number plate extraction has been compared with Kasaei
& Kasaei [7] and Azad & Shayegh [8] and is shown in
TABLE 1. Few sample images, which are successfully
recognized is shown in Fig.10.
Table 1: Results of number plate extraction method

Figure 7: Image obtained after ROI extraction

Figure 8: Image obtained after tilt correction

Figure 10: Few successfully recognized sample Images


www.ijsret.org

872

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 3, Issue 5, August 2014

3. RESULTS AND CONCLUSION


This is an efficient method for number plate extraction.
This method has been implemented on images with
different light conditions, different backgrounds,
different distances and low quality images. It provides
fast response to operation, and also correct tilt of number
plate with higher accuracy. By using this method the
correct extraction rate of number plate achieved has been
approximately 99.33% which is better than results
obtained by Kasaei & Kasaei [7] and Azad & Shayegh
[8]. In future other techniques can be used to further
increase the percentage efficiency of correct number
plate extraction.

Component, 3rd International Conference on Computer


and Knowledge Engineering, pp.21-25.
[9]
Database
is
taken
from
http://www.medialab.ntua.gr/research/LPRdatabase.html

REFERENCES
[1] D. Renukadevi, D. Kanagapushpavalli, (2011),
Automatic Licence Plate Recognition, 3rd International
conference on trendz in information sciences and
computing, pp. 75-78.
[2] Xiuxia Yu, Hongyu Cao, Haidong Lu, (2011),
Algorithm of Licence Plate Localization Based on
Texture Analysis, International Conference on
Transportation, Mechanical, and Electrical Engineering,
pp. 259-262.
[3] Rong-Tsai Lee, King-Chu Hung, (2012), Real-Time
Vehicle Licence Plate Recognition Based on 1-D
Discrete Periodic Wavelet Transform, International
Symposium on Computer, Consumer and Control, pp.
914-917.
[4] S. Qing-kun, Yuan Hui-jun, Zhou Teng, (2012),
Licence Plate Recognition Based On Mathematical
Morphology Method and RBF Neural Network,
international Conference on Measurement, Information
and Control, vol-2, pp. 782-786.
[5]
Ms.C.
Chandhok,
Mrs.SoniChaturvedi,
Dr.A.AKhurshid, (2012), An Approach to Image
Segmentation using K-means Clustering Algorithm,
International Journal of Information Technology (IJIT),
Vol-1, Issue-1, pp. 11-17.
[6] R. Timothy Marler, Jasbir S. Arora, (2010), The
weighted sum method for multi-objective
optimization: new insights, Struct Multidisc Optim,
vol-41, issue-6, pp. 853-862.
[7] S. H. M. Kasaei, S. M. M. Kasaei, (2011),
Extraction and recognition of the vehicle licence plate
for passing under outside environment, European
Intelligence and Security Informatics Conference, pp.
234-237.
[8] Reza Azad, Hamid Reza Shayegh,(2013), New
Method for Optimization of Licence Plate Recognition
system with Use of Edge Detection and Connected

www.ijsret.org

873

You might also like