International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 5, Issue 1, January 2015)
Automatic Vehicle Number Plate Recognition Using
Morphological Edge Detection and Segmentation
Teena Singh Rajput
Department of Computer Science and Engineering, Lala Lajpat Rai Institute of Engineering and Technology, Moga, India
Abstract- The vehicle number plate recognition
automatically controls access to a secured area for
authorized members. We tested our method for number
plate recognition which includes five main testimonials first
is pre-processing of the input image, then cropping the pre-
processed image that follows by the image edge extraction
operated on the cropped image after this the character
segmentation process algorithm and then finally to the
character recognition which gives the improved results that
shows our method has the average precision that can be put
into practice.
Keywords-- Number plate; Image edge extraction;
Vehicle number plate recognition; Character recognition;
Character segmentation.
I. INTRODUCTION
The automatic vehicle number plate recognition
portray in real-life applications to spot vehicles by
encapsulating their number plate. Several works analyze
this issue and specify many unfolding to allocate with
this work. This increasing passion for the improvement
of road safety in various areas like the automatic-toll tax
collection, rush hour law enforcement, parking transport
access control, road traffic supervision and with the
crime prevention [1-3]. It is executed with the synthesis
of five main procedure flow one after other in vehicle
number plate recognition method that are the pre-
processing of the query image, cropping the pre-
processed image then the edge extraction applied on the
cropped image that follows the character segmentation
and finally, the character recognition. Among this,
Segmentation is the most important part in the method
because it affect the system accuracy.
There are many issues to be fixed in order to recognise
a number plate successfully. The main issues while
capturing vehicle number plate is the variation in the
plate type and environment changes this cause challenges
in the recognition of the vehicle number plates. Their
main limitation is that number plate variations due to
plate types are the location, quantity, colour, and
occlusion, standard versus vanity, inclination and fonts.
The number plate is also limited due to the environmental
conditions like illumination in which the lighting and the
vehicle headlights and the other can be the background
which may have different patterns and textured floor.
500
Vehicle number plate recognition mainly consists [4]
of the five main testimonials. All the steps have their own
importance and methods followed in the number plate
recognition. In the first stage, pre-processing which
reduces the noise of the image and make it more clearly
visible. Next step will crop the image and then, the
number plate edge extraction is performed. This can be
done using many simple methods; it extracts the features
of the number plate as well as their shape and symmetry.
Then it will segment the characters individually to
recognise each character of the number plate using
algorithm which also gives the match degree results of
the each characters of the number plate.
The number plate determined in the pre-processing,
cropping the object and the edge-extraction is performed
under the classification stage. Next, the identification
stage, which involves the two main tasks that are the
character splitting or segmentation, and the character
recognition. The character splitting or segmentation is
main step in the whole process and even difficult, this
can be performed through various processes.
This include numerous different algorithm like the
genetic algorithm, the artificial neural networks, the c-
means fuzzy algorithm, the supportive vector-machine,
the markov processes algorithm, and the finite automata
processes. These methods can be widely specified with
the iterative approaches and the non-iterative approach.
II. RELATED W ORK
Data describing that the Car License Plate Recognition
has complex objective. This work was proposed by A.
conci et al. [5] to show a system that solves the process
of image acquition through optical character recognition
to achieve an automatic identification of plates.
Subsequently, the paper discussed a novel approach by
the M. Rama Bai et al. [6] for the removal of the noise as
well as the detection of the edges for both binary and the
gray-scale images using morphological operations. The
results demonstrate that filter cum edge detection
procedure overcomes the limitations of traditional
methods and efficiently clear the noise and extract more
prominent edges.
The current practise given by the Deepak Ghimire and
Joonwhoan Lee [7] proposed a method to enhance the
colour images with the non-linear transfer function by
retrieving the neighbouring pixels.
 International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 5, Issue 1, January 2015)
In this method, the enhancement is applied on the As, the closure and the interpretation shows from all
image where 'V' is the value of the luminance. The above work that were discussed in all the above papers
components are unchanged to restrict colour balance of that the combination of the morphological edge detection
the degradation between the HSV components. The and recognition with the correlation has not yet discussed
objective and subjective achievement evaluation shows so far for the vehicle number plate identification.
that the enhancement method that proposed and thus,
yields results positively when compared with III. P ROPOSED W ORK
conventional methods. To summarize the proposed approach. First, to
The improved approach was presented by Alyani propagate the input image I(x, y), hence, pre-processing
Ismail et al. in which the three main contributions VEDA is applied to the image which eliminates all the noise and
compared to sobel operator for veracity, algorithm the resultant image is . Then mathematical
complexity and time to process. Therefore, the CLPD morphological operations are applied on to get
meets the results for the real time requirements. David edge extracted image E(x, y). After this, the segmentation
Felguera-Martin et al. reported [8] the study for the is addressed to the acquired image to processes by
interferometry continuous modulated linear frequency applying character segmentation algorithm. Finally, the
wave to improve two different systems for the
character recognition algorithm is used to identify the
configuration of the radar. Unified approach was resultant image of the number plate.
implemented [9] in field-programmable gate array
(FPGA) built on peripherals component inter connect
(PCI)-format which is adaptable with the standardized Original Pre-processing of the image
computers that get objects from webbed cameras. Image I(x, y)
Therefore, through this system, the new vehicle is
determined. The images were correlated besides the huge
database with several huge number of references objects Cropping the image
to represent vehicle variations find in the regional
command. A real-time system correlates and performs
thousands of correlation per second which allow vehicle
identification timely that is less than 1s which was then
compared with the large number of reference images.
Morphology
The objective of paper is systematic survey of an
edge
existing ALPR research divide it to represent a view on
the state-of-the-art techniques. Hence, the features used extraction
for each stage were compared with the recognition
accuracy and the processing speed. An alternative
representation is the main contribution in this paper using
log-Gabor filter is proposed and evaluated by Jon
Character segmentation
Arrospide and Luis Salgado to design the descriptor
based on log-Gabor functions for vehicle verification.
The purpose of study is experimentally demonstrate the
referred theoretical superiority of log-Gabor filters over
standard Gabor filters to represent the better frequency Image Recognition Recognized
properties. The recent research in image segmentation
techniques addresses a various methods in this survey Object F(x, y)
[10] that analysis and observed that a hybrid solutions
consists being the best to solve the problem of image Fig. 1 Proposed work flowchart
segmentation. A. Pre-processing
Sukhwinder Singh and Neelam Rup Prakash [11] were
engaged in defining the edge detection by the novel The pre-processing of the number plate recognition is
algorithm that yields on multi-structure morphology the first step to remove the useless information from the
elements to get eight directions by using five edge input image I(x, y). The noise of the image should be
detection results. The opening, the closing, the top hat removed and the useful information should be acquired
transform and the bottom hat transform to get final by enhancing visual appearance of the image. When the
results using the edge detection by the morphological useless information is cleared then the greyscale contrast
algorithm and edge detection operator. enhancement improves the appearance by brightening the
dataset and the resultant image is .

501
 International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 5, Issue 1, January 2015)
The caution should be taken to the enhancement This is performed by using two basic morphological
techniques because this can lead to loss of information if operations that are dilation and erosion.
not correctly used.

(a)

(b)
Fig. 2 Original input vehicle number plate (a) plate1.jpg, (b)
plate2.jpg Fig. 4 Morphological edge detection applied on plate1.jpg (a)
Binarized images, (b) Morphological edge detection images, and (c)
B. Cropping Dilated images using morphologic operation.
Next, precisely containing the useful information from
the then by using the binary coordinates to crop
the vehicle number plate which is only required for
further processes that results in the C(x, y) image. At this
stage, the characters and numbers that were cropped from
the query number plate may also contain garbage objects
as well as useful objects.

(a)

(b)
Fig. 3 Cropped images of the vehicle number plates (a) plate1.jpg
cropped image, (b) plate2.jpg cropped image
This can be obtain by simply cropping the query
image only to the extend where the characters and the
numbers are visible to eliminate the unwanted region of
the query image.
C. The Edge Extraction using Morphological operations
First, the cropped image C(x, y) should be converted to
the binary image consists of only 1's and 0's (for black
and white). The edge of the image which is directly
extracted by the morphological basic operations gives the
E(x, y) from the cropped image C(x, y).
502
Fig. 5 Morphological edge detection applied on the plate2.jpg. (a)
Binarized images, (b) Morphological edge detection images, and (c)
Dilated images using morphologic operation
These two methods are based on the fundamental
morphological operations by giving the limits 1.1 and the
image terminals by selecting different size and extensions
converted to images of grey levels and then remove the
entire unwanted region. This process, first gives the
number of pixels connected together in a sequence to
form a group of connected objects. Then, it counts the
cropped characters and number of the connected region
from number plate and matches this characters and
number from the dataset of the template library which is
created to call the objects.
 International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 5, Issue 1, January 2015)
D. The Character Segmentation
(1)
After the edges of the number plates are determined as
E(x, y), the character segmentation is the next process in
the identification module. Each character of the cropped The value of CC lies between 0 and 1 where as the
number plate need to be split and the acquired image is value increases towards 1 will be the best correlation
the resultant image S(x, y). The character recognition value. The CC value calculated for the proposed
algorithms are many but the recent algorithm proposed is algorithm for the recognized image gives the results as
Otsu's algorithm. This algorithm identifies separate Table 1 shown below. This gives the approximation, as
objects within the image. It finds, that the region of the the true result according to all the causes of the image
connected pixels with similar properties and also finds kept in mind as through the illumination factor as well as
the boundaries between the regions. the environmental factors of the plate.
TABLE 1.
CORRELATION VALUES FOR EACH CHARACTERS OF
TWO INPUT NUMBER PLATES

FRENCH 4U YOUR REG

(a) (plate2.jpg) (plate2.jpg)
Characters Correlation Characters Correlation
values values
F 0.5602 Y 0.6307
R 0.5365 O 0.7305
E 0.6566 U 0.7654
(b)
N 0.5377 R 0.5248
Fig. 5 Segmented number plates characters (a) plate1.jpg, (b) C 0.7297
plate2.jpg
H 0.7333 R 0.5248
E. Character Recognition
4 0.5349 E 0.6934
The final recognition stage is performed by the
correlation matching algorithm implementation and the U 0.8704 G 0.5832
accuracy rate is the approximation of value 1 in the F(x,
y) image. This final resultant recognized objects of the V. CONCLUSION AND FUTURE SCOPE
vehicle implements correlation for the pattern recognition
by the following steps: This paper presents a combination of both theoretic
practise and the visual results on the vehicle number plate
i. it sequentially multiplies every segmented objects
recognition through the hybridization of the
S(x, y) over the entire reference image set from the
morphological edge detection and proposed Otsu's
template library,
algorithm for the segmentation. The approximate results
ii. after this, it calculates the correlation plane for each
show that this approach gives the potential of the
multiplication,
morphological edge detection and segmentation for
iii. it identifies the maximum peak value as max(out) to
vehicle number plate classification that has been
the correlation for each segmented character S(x, y),
assessed. This method have better theoretical properties
iv. it sorts the maximum peak correlation values from
than the traditional filters that represent images but still
each correlation plane to found the reference image
not used for vehicle verification. The five main modules
brings about the highest correlation peak value, i.e.,
that worked together one after one for the improved
with the reference image S(x, y) have the best match
results even lead to the satisfaction to the natural moving
with unknown input image I(x, y).
vehicle number plates. The purpose to apply this five
testimonial approach is that to eliminate the unwanted
IV. RESULTS AND D ISCUSSIONS
region from the vehicle number plate and detect the
The correlation matching algorithm gives the absolute prominent edges which are required for the recognition
values of the match degree of the characters for the two of the characters and numbers from the vehicle number
query images namely plate1.jpg and plate2.jpg taken plates. Hence, the recognition accuracy is high. The
from the matlab library of 512*512 sizes. The results method is quite simple that can be put into practise and
were tested and validated using correlation coefficient also has a good application prospect.
(CC) metrics [12]. CC is used to match the characters The future scope can be extent to the process that may
from the template library and the vehicle number plate. work with the character recognition as the mobile
application.

503
 International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 5, Issue 1, January 2015)
REFERENCES
[1] M. Ibrahim and M. Shehata, Automatic license plate recognition
(ALPR): A state-of-art review, IEEE Transaction on Circuits and
Systems for Video Technology, Vol. 23, NO. 2, February 2013.
[2] J. Arrospide and L. Salgado, "Log-gabor filters for image-based
vehicle verification," IEEE Transaction on Image Processing, Vol.
22, No. 6, June 2013.
[3] A. M. Al-Ghaili, S. Mashohor, A. R. Ramli and A. Ismail,
Vertical edge-based car license plate detection method, IEEE,
Manuscript Received January 25,2012.
[4] C. Chenyu, C. Baozhi, C. Xin, W. Fucheng, Z. Chen,
"Application of image processing to vehicle license plate
recognition," Proceeding to 2nd International Conference on
Computer Science and Electronics Engineering (ICCSEE 2013).
[5] A. Conci, J. E. R. de. Carvalho and T. W. Rauber, "Graphical
models for joint segmentation and recognition of license plate
characters, "IEEE, Latin America Transactions, Vol.7, No. 5,
September 2009.
[6] M. R. Bai, V.V. Krishna, and J. Sreedevi, "A new morphological
approach for noise removal cum edge detection," IJCSI
International Journal of Computer Science Issues, Vol. 7, Issue 6,
November 2010.
504
[7] D. Ghimrie and Joonwhoan Lee, "Nonlinear transfer function-
based local approach for color image enhancement," IEEE
Transaction on Consumer Electronics, Vol. 57, No. 2, May 2011.
[8] D. F. Martin, J. T. G. Partida, P. A. Gonzalez and M. B. Garcis,
"Vehicular traffic serveillance and road lane detection using radar
interferometry," IEEE Transaction on Vehicular Technology, Vol.
61, No. 3, March 2012.
[9] G. J. McDonald, J. S. Ellis, R. W. Panney and R. W. Price, "Real-
time vehicle identification performance using FPGA correlator
hardware," IEEE Transaction on Intelligent Transportation
Systems, Vol.13, No. 4, December 2012.
[10] W. Khan, "The image segmentation techniques: A survey,"
Journals of image and graphics, Vol. 1, No. 4, December 2012.
[11] S. Singh and N. R. Prakash, "Edge detection of grey scale images
based on multi-structure elements morphology, "International
Journal of Latest Research In engineering and Computing
(IJLREC), Vol. 1, pp. 48-51, November-December 2013.
[12] A. Mahmood and S. Khan, "Correlation coefficient based fast
template matching through partial elimination," IEEE Transaction
on Image Processing, 11 August 2010.