See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/265991428

Number Plate Recognition for Indian Vehicles

Article

CITATIONS READS

23 1,508

2 authors, including:

Sandipan Pralhad Narote

Sinhgad Technical Education Society
33 PUBLICATIONS   228 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Sandipan Pralhad Narote on 16 January 2015.

The user has requested enhancement of the downloaded file.

IJCSNS International Journal of Computer Science and Network Security, VOL.11 No.2, Feb. 2011 143

Number Plate Recognition for Indian Vehicles

M. M. Shidore†, S. P. Narote††

†
Vishwakarma Institute of Technology, University of Pune, Pune, India
††
Sinhgad College of Engineering, University of Pune, Pune, India

Abstract: This paper presents Automatic Number Plate features of characters, each character is recognized, in the
extraction, character segmentation and recognition for character recognition phase. Extraction of number plate is
Indian vehicles. In India, number plate models are not difficult task, essentially due to: Number plates generally
followed strictly. Characters on plate are in different occupy a small portion of whole image; difference in
Indian languages, as well as in English. Due to variations number plate formats, and influence of environmental
in the representation of number plates, vehicle number factors. This step affects the accuracy of character
plate extraction, character segmentation and recognition segmentation and recognition work. Different techniques
are crucial. We present the number plate extraction, are developed for number plate extraction.
character segmentation and recognition work, with english Hao Chen et al [1] proposed the method in two steps.
characters. Number plate extraction is done using Sobel First, several candidates based on texture information
filter, morphological operations and connected component similar to number plate are extracted. In the second step,
analysis. Character segmentation is done by using auto-correlation based binary image and projection
connected component and vertical projection analysis. algorithm are used to verify the true candidate plate. Gisu
Character recognition is carried out using Support Vector Heo [2] developed number plate detection algorithm using
machine (SVM). The segmentation accuracy is 80% and group of lines forming rectangle at the plate boundary.
recognition rate is 79.84 %. Followed by this step is the vertical edge density algorithm
Key words: to find out the plate area. Ozbay et al [3] devised smearing
Vehicle number plate, Number plate extraction, character algorithm to locate the number plate. In Korean license
segmentation, character recognition. plate extraction, Mei Yu et al [4] proposed vertical edge
detection followed by size, shape filter for edge area and
1. Introduction edge matching technique based on plate model. Farhad
Faradji et al [5] first used Sobel vertical edge detection on
Automatic Number Plate Recognition (ANPR) system is the image. Next, vertical projection analysis was used to
an important technique, used in Intelligent Transportation locate plate area. Fake candidates were deleted using
System. ANPR is an advanced machine vision technology compact factor, which evaluated the densest vertical edge
used to identify vehicles by their number plates without area declaring true number plate.
direct human intervention. It is an important area of Every character on detected number plates is
research due to its many applications. The development of segmented in character segmentation step. Segmentation
Intelligent Transportation System provides the data of techniques based on projection analysis, Hough transform,
vehicle numbers which can be used in follow up, analyses region growing are proposed in the literature. Xinagjian
and monitoring. ANPR is important in the area of traffic He et al [6] used horizontal and vertical projection analysis
problems, highway toll collection, borders and custom for character segmentation.
security, premises where high security is needed, like Yuangang Zhang et al [7] implemented character
Parliament, Legislative Assembly, and so on. The segmentation using Hough Transform. In this, horizontal
complexity of automatic number plate recognition work edges of the plate area were decided initially, using Hough
varies throughout the world. For the standard number plate, Transform, which helped to segment the characters with
ANPR system is easier to read and recognize. In India this the large rotation. Characters were segmented using
task becomes much difficult due to variation in plate vertical projection analysis based on the prior knowledge
model. of the plate model. Feng Yang et al [8] developed region
The ANPR work is generally framed into the steps: growing algorithm for character segmentation. Shen Zheng
Number plate extraction, character segmentation and Wang et al [9] used connected component analysis for
character recognition. From the entire input image, only character segmentation.
the number plate is detected and processed further in the Fundamental issue in character recognition is shape
next step of character segmentation. In character analysis. For the recognition of segmented characters,
segmentation phase each and every character is isolated algorithms use Pattern/ Template matching, Computational
and segmented. Based on the selection of prominent Intelligence, statistical classifiers.

Manuscript received February 5, 2011

Manuscript revised February 20, 2011
144 IJCSNS International Journal of Computer Science and Network Security, VOL.11 No.2, February 2011

ANPR system consists of three modules; 1) Extraction

of number plate, 2) Character segmentation, and 3)
Character recognition. In this paper ANPR wok for Indian
car is presented. Images are taken out with different
illumination conditions, different background and
(a) (b) (c)
orientation. Histogram equalization, median filter are used
which take care of lighting and contrast problem. Sobel Fig. 1: (a) Gray scale image (b) Sobel vertical edge detection (c) Effect of
vertical edge detection and morphology is employed to threshold
locate the number plate. Projection analysis is used to
segment the characters present on the plate. For 2.3 Candidate Plate Area Detection
recognition work SVM is used.
The rest of the paper is organized as follows: Section 2 Morphological operations aim to remove unrelated objects
explains the proposed algorithm for number plate in the image. Dilation and erosion are used to extract
extraction; Section 3 explains the algorithm used for candidate plate areas from the entire image. Sometimes
character segmentation; Section 4 describes the number background areas may also get declared as candidate plate.
plate recognition algorithm using SVM. Experimental Hence to remove the fake candidates, plate validation is
results are demonstrated in Section 5. Section 6 concludes done using the aspect ratio of the plate and horizontal cuts
the paper. [11] in the number plate.

2. Number Plate Extraction 2.4 True Number Plate Extraction

Number plate extraction is the key step in ANPR system, After the detection of candidate number plate area,
which influences the accuracy of the system significantly. Bounding Box analysis is used to extract plate area from
The goal of this phase, given an input image, is to produce the original image. From the Bounding Box analysis,
a number of candidate regions, with high probability of respective row and column indices of plate area are found
containing number plate and validate for true number plate. out. Once the indices of number plate are known, the
number plate is extracted from original gray scale image.
2.1 Image Acquisition and Preprocessing The result is as shown in Fig. 2.

In this system a high resolution digital camera is used to

acquire an image. Images are taken in different
background, illumination conditions, and at various Fig. 2: Extracted number plate
distances from the camera to vehicle. Images are resized to
(1024 X 768). All the processing steps are executed on 3. Character Segmentation
gray scale image. Preprocessing is mainly used to enhance
Character isolation from the number plate region is the
the processing speed, improve the contrast of the image,
important step in ANPR system, which influences the
and to reduce the noise in the image. In order to reduce the
accuracy of character recognition significantly. The goal of
problem of low quality and low contrast in car images,
this phase, given the number plate image, is to segment all
images are enhanced by using histogram equalization on
the characters, without loosing features of the characters.
gray scale image.
This phase consists of the sequence of operations as,
character region enhancement, connected component
2.2 Vertical Edge detection analysis and projection analysis.
The characters on number plate region contain abundant
3.1 Character Region Enhancement
edges as compared to background area. This feature is
employed for locating the candidate plate area from the
To enhance the character region, if histogram equalization
input image. Sobel vertical edge detection is used to find
or gray level scaling is used, noise may also be enhanced.
out the regions which have high pixel variance value [10].
For the correct character segmentation, only the character
To extract candidate number plate area from the entire
pixels need to be enhanced and the background pixels need
image, threshold is used to select rows which are having
to be weakened. For this ‘gray threshold’ [10] is
particular white pixel density. Fig. 1 shows the result of
implemented. Fig. 3 shows thresholding effect.
effect of using, Sobel vertical edge detection and
threshold.

Fig. 3: Binarized number plate

3.2 Connected Component Analysis
IJCSNS International Journal of Computer Science and Network Security, VOL.11 No.2, February 2011 145

To remove the noise, other than characters on the plate, utilization time, particularly when the features are
connected component analysis is used. Each labeled redundant.
matrix of 8-connectivity pixels is evaluated based on the Initially, the centroid of the character image is
area threshold. The result of connected component determined. With respect to centroid, number of
analysis and noise removal is as shown in Fig. 4. transitions along the axes, 0 to 1 and 1 to 0, up to the
boundary of character are counted. Transitions are
specified for axes with predetermined angles. Fig. 6 shows
the nature of axes towards boundary, counting number of
(a) (b) transitions. By observing strokes of each character, 13
Fig. 4: (a) Connected component analysis (b) Noise removed different angles are decided to count the transitions.

3.3 Vertical Projection Analysis

Character segmentation is implemented to isolate each

character on the number plate. Vertical projection analysis
is used to find the gaps between the characters. The
characters are segmented based on the number of valleys Fig. 6: Input character image overlaid with nature of axes
in projection. Row and column indices of each character
are recorded, and it is extracted from the original gray 4.3 Number Recognition
scale number plate. The vertical projection and segmented
characters are shown in Fig. 5. This proposed work The character recognition work is carried out using SVM
isolates every letter and digit on the number plate oriented as classifier. The car number plate consists of combination
horizontally in one row, along the width of number plate. of 0 to 9 digits and alphabets of A to Z. All the digits as
well as alphabets are assigned a class label. For multi-class
classification, either one-against-one approach is used or
one-against-all is preferred. In the proposed work
recognition is achieved using one-against-all approach
[13]. The polynomial kernel is used for classification. The
database consists of 250 images. 60 % of the data is used
for training and remaining 40% is used for testing.

5. Experimental Results and Performance

Fig. 5: Vertical projection analysis and respective segmented characters
Analysis
4. Character Recognition
A database consists of different sized JPEG colored
images. Images are resized to 1024 X 768. Total 250
The character recognition phase consists of two steps: 1)
images are used to test the algorithm. The images are
Character normalization and feature extraction, 2)
taken with different background as well as illumination
Character classification using Support Vector Machine
conditions. Experiments show that the algorithm has good
[12].
performance on number plate extraction, and character
segmentation work. It can deal the images correctly, with
4.1 Character Normalization noise, illumination variance, and rotation to ±50. This work
is implemented using MATLAB.
Segmented characters have very much variation in size. In Table 1 illustrates number plate extraction and
this phase, all the characters are normalized to predefined character segmentation success rate.
height (Vertical Length) in pixel. As the characters always
have variable width (Horizontal Length), each character Table1 Performance of number plate extraction and character
image is normalized to a size of 32 X 32, by image segmentation.
mapping technique. Total No. Success
Algorithm
of Images Rate (%)
Number Plate
4.2 Feature Extraction Extraction
250 85
Character
214 80
The goal of feature vector is to define distinguishing Segmentation
features of the characters. Selecting the most relevant
feature of each character can not only facilitate data Deep shadows and reflections have an impact on number
visualization and data understanding, but also reduce the plate extraction work. Because of uneven illumination,
measurement, storage requirements, training and stained number plates, true number plates could not get
 146 IJCSNS International Journal of Computer Science and Network Security, VOL.11 No.2, February 2011

correctly extracted. Failure in character segmentation was [10] Rafel C. Gonzalez, Richard E. Woods, “Digital Image
mainly because of merging of characters on number plate, Processing,” 2nd edition, Prentice Hall, Inc., 2002.
stained number plates, orientation of the image and poor [11] Tran Dur Duan, Duong Anh Duc, Tran Le, Hong Du,
“Combining Hough transform and contour algorithm
illumination.
for detecting vehicle license plate”, Proc. of
Character recognition work is done on 10 digits (0 to International Symposium Intelligent Multimedia, Video
9) and 26 alphabets (A to Z). The recognition rate achieved and Speech Processing , 2004, pp. 747-750.
using ‘polynomial’ kernel is 79.84 %. The false [12] Christopher J. C. Burges, “A tutorial on Support Vector
recognition is due to similarity in the character shape, e.g. Machine for pattern recognition,” Bell Laboratories,
6 and B, 5 and S etc. Recognition can be increased if two Lucent Technologies, 1998.
groups of SVM’s are designed separately, one for digits [13] S. R. Gunn, “Support Vector Machines for
and other for alphabets. classification and regression, technical report,”
University of Southampton, 1998.
6. Conclusions
An algorithm for vehicle number plate extraction, Mrunal M. Shidore was born in
character segmentation and recognition is presented. India in 1964. She received the
Database of the image consists of images with different B.E. (Electronics) degree from
size, background, illumination, camera angle, distance etc. University of Pune, Pune in 2003.
The experimental results show that, number plates She is currently working as a
are extracted faithfully based on vertical edge detection Lecturer in Electronics
and connected component algorithm, with the success rate Engineering, with the
of 85%. Character segmentation phase using connected Vishwakarma Institute of
component analysis and vertical projection analysis works Technology, Pune. Her research and teaching interests
well with the success rate of 80%. The success rate include image processing.
achieved for character recognition is 79.84%.
Sandipan P Narote was born in
References India in 1975. He received the
B.E.(E&TC), M.E.(Elect.) from
[1] Hao Chen, Jisheng Ren, Huachun Tan, Jianqun Wang, Dr. Babasaheb Ambedkar
“ A novel method for license plate localization”, 4th University Aurangabad and
Proc. of ICIG 2007, pp. 604-609. Ph.D.(E & TC) from Swami
[2] Gisu Heo, Minwoo Kim, Insook Jung, Duk Ryong Lee, Ramanand Tirth Marathwada University Nanded in 2000,
Il Seok Oh, “Extraction of car license plate regions 2002 and 2010, respectively. He is currently a Professor of
using line grouping and edge density methods”, Electronics and Telecommunication Engineering with the
International Symposium on Information Technology Sinhgad College of Engineering, Pune, India. His
Convergence, 2007, pp. 37-42. research and teaching interests include signal processing,
[3] Serkan Ozbay, Ergun Ercelebi, “Automatic vehicle
image processing and biometrics.
identification by plate recognition”, Proc. of PWASET,
vol. 9, no. 4, 2005, pp. 222-225.
[4] Mei Yu and Yong Deak Kim, “An approach to
Korean license plate recognition based on vertical edge
matching”, IEEE International Conference on System,
Man and Cybernetics, 2000, vol.4, pp. 2975-2980.
[5] Farhad Faradji, Amir Hossein Rezaie, Majid Ziaratban,
“A Morphological based License Plate Location”, ICIP,
2007, pp. I 57-I 60.
[6] Xiangjian He et al, “Segmentation of characters on car
license plates”, 10th Workshop on Multimedia Signal
Processing, Oct. 2008, pp. 399-402.
[7] Yungang Zhang, Changshui Zhang, “A New algorithm
for character segmentation of license plate”, Proc. of
IEEE Intelligent Vehicles Symposium, 2003, pp.
106-109.
[8] Feng Yang, Zheng Ma, Mei Xie, “A Novel approach
for license plate character segmentation”, ICIEA, 2006,
pp.1-6.
[9] Shen-Zheng Wang, His-Jian Lee, “Detection and
recognition of license plate characters with different
appearances,” Proc. of 16th International Conference on
Pattern Recognition, vol.3, 2003, pp. 979-983.

View publication stats