International Journal of Scientific Research in Computer Science, Engineering and Information Technology

© 2020 IJSRCSEIT | Volume 6 | Issue 3 | ISSN : 2456-3307

DOI : https://doi.org/10.32628/CSEIT2063208

A Review Paper on Automatic Number Plate Recognition System

Shally Gupta*1, Rajesh Shyam Singh2, H.L. Mandoria3
*1 Department of Information Technology, College of Technology, Govind Ballabh Pant University of
Agriculture and Technology, Pantnagar, U.S. Nagar, Uttarakhand, India
2Department of Information Technology, College of Technology, Govind Ballabh Pant University of
Agriculture and Technology, Pantnagar, U.S. Nagar, Uttarakhand, India
3Department of Information Technology, College of Technology, Govind Ballabh Pant University of
Agriculture and Technology, Pantnagar, U.S. Nagar, Uttarakhand, India
ABSTRACT

Number plate recognition brings a drastic improvement for the city traffic enhancement. It provides the
direction in which the steps should be taken for working of an effective intelligent transportation system.
ANPR have become necessity for traffic control management due to rapid increment of vehicles. The main aim
of ANPR is to monitor traffic and for security purpose. Recognition of number plate uses image processing
techniques and latest technology in detecting characters on vehicle license plates automatically. In recent years,
there are many technological developments in recognizing license plate area of research. Image processing
protocols like OCR technology allow the traffic surveillance to deal with several problems that occurs in
criminal investigation, toll collection, monitoring traffic, controlling speed, parking management etc. For
efficient management of traffic and mass surveillance in transportation system, an ANPR system is essential.
With the aid of image processing algorithms and vehicle images dataset, it becomes possible to monitor traffic
at a large scale. Vehicle images are helpful for recognizing characters on license plates by performing image
segmentation, feature extraction and character recognition. Data collected through captured images are utilized
in the commercial applications, law enforcement, traffic applications etc. The software examines the vehicle
picture as an input image that results in displaying the plate numbers. The system with image processing used
reliably for traffic detection where modification of the technologies enables an accurate acknowledgement of
vehicle number plates. There are several ANPR systems developed, working on character recognition of LP by
the help of image processing technique. This paper reviews the performance by researchers in this particular
area towards meeting goals of transportation system. It also provides major issues and challenges in this field.
Keywords : Automatic Number Plate Recognition (ANPR), Optical Character Recognition (OCR), License Plate
(LP)

I. INTRODUCTION control vehicles traffic in present time [9]. For this

purpose, ANPR becomes a key factor for a quick
Vehicles perform a vital role in an element of traffic cognitive transportation system. In addition, in
control structure. It creates a huge problem for traffic present scenario, application complexity increases
control area to manage this large traffic. This is due to which in return demands the higher requirements for
rapid increase of population that is not easy to an efficient license plate recognition protocol [7].

CSEIT2063208 | Accepted : 20 June 2020 | Published : 28 June 2020 | May-June-2020 [ 6 (3) : 955-966 ]
955
 Shally Gupta et al Int J Sci Res CSE & IT, May-June-2020; 6 (3) : 955-966

For an efficient monitoring and management on difficulty due to these traffic violations. This may
roads, an Automated Number Plate Recognition include frauds in collecting of tolls on highways and
(ANPR) system plays a vital role. ANPR uses an parking lots, speeding issues, theft of cars and so on.
image processing algorithms, which help to detect Therefore, now a day the necessity of vehicle plate
the conveyance license plates, and recognizing identification is in demand. The information, which
characters of number plates. In a perfunctory is originated from the identification of vehicle
surveillance of pedestrians or vehicles management, numbers, can be assisted in miscellaneous purposes
an ANPR system helps to capture the vehicle like dealing with several crimes, monitoring of
possessor record by identifying there vehicle plates border crossing, highway toll stations, searching for
[9]. As ANPR is a bulk monitoring method, thus it stolen vehicles, access and flow control, proper traffic
plays an imperative role in traffic control protocol by management etc. [12]. ANPR also plays a vital role in
simply recognition of number plates. Here the smart cities for criminal investigation and traffic
pictures are clicked by camera and conveyance and detection and vehicle tracking. In manifold highway
its possessor is identified using of number plate toll systems observations were that there are fixed
recognition methods [1]. shooting angles and lightening environments, size
In 1976, the first ANPR system was designed in and brightness. Therefore for this aim, ANPR is used
United Kingdom at Police Scientific Development frequently in highway toll systems [6].
Branch. Although it’s main prototype system started Further advantages of ANPR system includes
operating after 1979 in which the contracts were monitoring of administration, urban road monitoring,
given to manufacture in industries. At very first outing vehicles are tested easily, checking of illegal
contract, it was working at EMI, Electronics and later entries, incidents managements, various safety
on it was given to the Computer Recognition Systems managements, used in surveillance of petrol posting
(CRS) in Workingham, UK [1]. forecourt, in an area of red light camera, customer
credentials, speed enforcements, crossing of border
Primary Applications of ANPR and so on [9].
The primary applications of number plate recognition
are used in urban traffic flow monitoring, detection Goals of ANPR- The detection of license plate is a
of vehicle seed at highway, auto transport charge, in very good source of information for record detection
traffic accessibility [8]. Other than this, observations and recognition. However, the traditional procedure
were found that the ANPR has the big data of number plate identification is tedious [7]. The
applications too i.e. the recovering of stolen cars, manual way of identifying the vehicle and its owner
automatic toll collections, parking management, is not that applicable in detecting license plates to
monitoring of road traffic, traffic law violation of retrieve the secret treasures of knowledge. Automatic
rules [4]. It is also utilized in several different fields license plate recognition is therefore chosen as an
like entrance admission, security control, airports, optimal solution for extracting numeral plate
speed control, harbour cargo control and many more information. Moreover, for this we require different
[5]. recognition techniques for the extraction of
information of number plates.
With the rapid growth in population, there is rapid
increase in vehicles too, which in return also increase How Number Plates are recognized – The plate
the traffic violations. When there is more public recognizing is performed in two modules:
traffic, the transport management system face

Volume 6, Issue 3, May-June-2020 | http://ijsrcseit.com

956
 Shally Gupta et al Int J Sci Res CSE & IT, May-June-2020; 6 (3) : 955-966

• License plate detection A- Cost of ANPR System

• License plate recognition There are various types of ANPR that are available
and due to different types, component, uses and
The basic steps involved in identifying vehicle geographical areas they fall into different categories
number plates are image acquisition or input image, of price. It depends on the authority criteria for
pre-processing of image involves gray scale which they wish to purchase ANPR program system.
conversion and noise reduction, segmentation of In market of 2016 evolution, it was around USD 1.78
character, extraction of features, classification and Billion. It is highly anticipated by the end of 2023,
identification of character. the price may increase up to USD 3.57 Billion.

B- Features
All the ANPR contain a camera that differs in size
and quality that totally depends on the price of
ANPR. The pictures taken from camera can be from
infrared, colour or both mixtures. The system
sometimes consists of a chain of frames that affects
the accuracy rate. The cameras of ANPR are consists
of IR-lamp, fast shutter and zooming control of lens
Fig 1. Basic steps in ANPR System
for detecting images. For achieving better plate
information, approximately 40 pixels of width are
In this paper, the division is in 6 sections. Section 2
required. The distance of camera from the
provides a brief introduction of features and uses of
conveyance should be as minimum as possible that is
ANPR. Also, depicts advantages and disadvantages.
approximately <35 meters. The vertical angle of
Section 3 throws light on brief literature review of
camera on license plate should be < 45 degrees, and
work done by other researchers. Section 4 provides
the horizontal angle of camera should be < 45 degrees.
comparative study and section and highlights the
It has observed that the minimum 2MP camera
important issues and challenges. Finally, in Section 5
required for parking lot management and collecting
conclusion is provided.
tolls automatically, 4MP camera for logistics or
vehicle repair type works and approximately 8MP for
II. FEATURES OF ANPR
highway traffic monitoring camera is used.

License plate recognition is achieved through device

C- Powerful Source
program ANPR. Images taken by using cameras
It completely relies on the pace and camera angle of
displays the output of vehicle plate. ANPR's basic
the vehicle to achieve an accurate accuracy for
work is reading and unfolding license plate. ANPR
recognition. Camera measures this by capturing
are also called by the name of ALPR (Automatic
innumerable vehicle images and further re-
License Plate Recognition). The software of this
examining those images for better accuracy.
system forms a record of all number plates that
interpret with all the related data such as date, time
and GPS location. It utilizes the OCR technology for
D- Ease and Use
recognizing characters of number plates.
In this system, the images of vehicle pictures are
being stored therefore, ANPR needs huge bandwidth.

Volume 6, Issue 3, May-June-2020 | http://ijsrcseit.com

957
 Shally Gupta et al Int J Sci Res CSE & IT, May-June-2020; 6 (3) : 955-966

Therefore, to decrease the storage, images are being • For helping of visitor management propagation in
converted to text and only the record of plate respect to detect outside vehicles.
numbers are maintained followed by date. • ANPR are efficiency saving by reducing manual
traffic check and access permission to required
Use of ANPR in smart city members only.
The transportation in miscellaneous areas has • Extra security is provided by installing ANPR
resumed being unbelievably burdensome. Therefore, such as reduced human error. Unauthorized
violation of regulations is common now days. There vehicles are located easily.
are elevated pace locating vehicles cameras that aid in • Efficient way for resource management to detect
detecting conveyance speed and finding traffic the in and out time of an employee can be done
violation by any vehicle by keeping the record of by using ANPR.
number plate. • The vital role in parking management is granted
by ANPR system by controlling the traffic flow.
They are utilized in the technologies such as image
processing, communication purpose, sensor, control Disadvantages ANPR
etc. [8] These technologies provides insight and
details about the vehicle regarding the vehicle owner, • ANPR creates problem in extreme weather
law enforcement etc. and hence their use provides conditions.
constant, timely and effective help in transport • Privacy issues are a significant element in ANPR,
management. as sensitive information is collected in the system
and can be misused further.
Since the manual recording of vehicle, information • Photos cannot be identified because of the
may be recorded wrong and the chances of blurring of the picture while the car is going by.
information breach and misplaced data are higher. • The photos with low resolution often lower the
Therefore automatic software program for traffic accuracy rate in the ANPR method.
detection is needed for the area of safety purposes
and for many other traffic tasks [3]. III. REVIEW LITERATURE

ANPR plays a major role in the age of Big Data Saqib Rasheed et. al [2012] proposed “Automated
deployment. In a complex environment conditions number plate recognition using hough lines and
like ambient lightening the program is effective to template matching” proposed a robust license plate
use [7]. identification and recognition system based on hough
lines using Hough Transformation and template
ANPR collect vehicle images as data, displays plate matching Model. In this research license plate
numbers by the aid of algorithms, and depicts the detection is carried out using canny detector and
useful information. Now-a-days plate recognition transformation of the hough. Additionally, the
systems are a fantastic tool with the aid of cameras identification of license numbers is achieved using
for capturing transport images. matching templates. The accuracy rate for vehicle
plate extraction is 94.11 percent and the number of
Advantages of ANPR vehicles recognized successfully is 90.62 percent. The
proposed ANPR system is a standardized number

Volume 6, Issue 3, May-June-2020 | http://ijsrcseit.com

958
 Shally Gupta et al Int J Sci Res CSE & IT, May-June-2020; 6 (3) : 955-966

plate for Islamabad, which are 89.70 percent overall of morphological images. As a feature extractor, a
results. [01] pre-trained Convolution Neural Network (CNN)
"Alex-Net" is used, and SVM is used as a classifier.
Alice N Cheeran et. al [2017] proposed “Automatic The algorithm is implemented in morphological
Car Number Plate Recognition” have used Modified image processing in C++ programming language, and
Ant Colony Optimization Algorithm in the plate in pre-trained CNN and SVM sections in MATLAB.
localization for the edge detection. The Kohonen The proposed device has been checked on 126 sample
neural network is used to classify an individual images, including 113 extracted as the right number
character's location and dimensions. In addition, the plate. It gives 89.7 per cent success rate. [4]
connected component analysis (CCA) and the
proposed hybrid network are compared using CCA, Dinesh Bhardwaj et. al [2014] proposed “Automated
and the neural kohonen network is conducted and Number Plate Recognition using Machine Learning
the accuracy is measured. Eventually, a hybrid Algorithms” have proposed a technique to recognize
hierarchical classification system based on inductive a number plate of vehicle. In this study, the
RULES-3 approach and SVM approach is proposed. recognition of number plate was rendered in five
The proposed character segmentation and extraction separate steps, such as Image Acquisition with the aid
algorithm gives 94 percent accuracy. [2] of a high-resolution camera, pre-processing of image
using gray scale conversion, removal of noise using
Sumanta Subhadhira et. al [2014] proposed “License median filter, edge detection using Sobel Technique,
Plate Recognition Application using Extreme segmentation using vertical technique, character
Learning Machines” have developed new license recognition using K-Star machine learning algorithm.
plate recognition system using Extreme Learning Here dataset was collected as 320 images of various
Method (ELM). Here smartphone application is used vehicles in real-times, parking etc. This work has
to photograph the license plate of a vehicle. This been performed for the Indian ANPR system. [5]
system recognizes the Thai license plate to pre-
process and extract features using ELM as a classifier Wei-chen Liu et. al [2017] proposed “A Hierarchical
and a Histogram of Oriented Gradients (HOG). For Plate Recognition System using Supervised K-means
this, the program follows the same concepts as and Support Vector Machine” have presented a novel
provincial license plate recognition. ELM is used for hierarchical character recognition scheme based on
data concerning segmentation. The province part is supervised K-means and SVM to recognize the
95.05% accurate but it has less data to process. The blurred and tilt license plate. In this proposed work,
proposed application can recognize the whole plate the main aim was to reduce classes of characters in
which consist both the car registration identification each subgroup and then further reduce the numeral
number and it’s province by 89.05% accuracy. [3] of SVMs and their complexity. This system offers
98.89 percent precision for blurred license plate and
Yoshihiro Shima [2016] proposed “Extraction of tilt characters. Further on the comparison with the
Number Plate Images Based on Image Category state-of-the-art of plate recognizing approaches, the
Classification using Deep Learning” has proposed the average improvement is found to be 3.6%. [6]
ANPR system on a region extraction technique
combined with morphological image processing and Yongsheng Li et. al [2019] proposed “Vehicle License
deep learning. Edge detection and Connected Plate Recognition Combining MSER and Support
Component Analysis (CCA) is used in the processing Vector Machine in a Complex Environment” have

Volume 6, Issue 3, May-June-2020 | http://ijsrcseit.com

959
 Shally Gupta et al Int J Sci Res CSE & IT, May-June-2020; 6 (3) : 955-966

proposed a license plate recognition algorithm based extraction is done on the basis of six zone features.
on MSER and SVM. Here the license plate For the classification of characters and numbers on
recognition algorithm is contrasted with two the number plate region are performed in WEKA
standard license plate recognition algorithms i.e. edge tool. Detailed accuracy is done by using K-star ML
detection and detection of colours. The proposed algorithms which show the accuracy on data set of
algorithm is suitable for complex surroundings. 99%. All over it is observed that among all K-star
MSER extracts the character area directly. Here in algorithms performs better for recognizing number
MSER, license plate location and segmentation is not plate. [9]
carried out and directly screened for the characters
which is directly used for the character recognition. Mahima Satsangi et. al [2018] proposed “License Plate
This composite MSER and SVM system gives more Recognition: A Comparative study on Thresholding,
than 90 per cent accuracy. And the edge detection OCR and Machine Learning Approaches” have
and color detection accuracy is poor as compared to discussed license plate recognition using Viola Jones
the MSER and SVM combined. [07] algorithm. In this the main concentration was on the
classification and the recognition of characters on the
Yang Guang [2011] proposed “License Plate license plate. The images captured are with the aid of
Character Recognition Based on Wavelet Kernel LS- magnetic loop detector sensor. The identification of
SVM” has presented license plate character based on the license plate is performed essentially in three
wavelet LS-SVM (least square SVM). The Mexico hat sequential steps such as image detection, number
wavelet kernel is used in this work because, due to plate extraction and image segmentation. The ML
the function of multi-scale analysis, it has better algorithm viola Jones is used which consist of four
generalizing potential and is approximately basic concepts like Haar feature, Integral images,
orthogonal. The approach based on the LS-SVM Ada-boost and a Cascade classifier. The output
wavelet kernel consists of two key steps in character obtained from the algorithm proposed is compared
recognition which are preprocessing and multi- with the output obtained from threshold and OCR
classifier. Here, according to BT-model, four technology. The accuracy from the all three
classifiers are based on wavelet kernel LS-SVM. The implementations shows that the viola jones gives the
experiment is performed on the BP neural network best performance of the 80%. [10]
and RBF LS-SVM method proposed for this method.
The system suggested has a cumulative recognition Kaili Ni et. al [2018] proposed “A Proposed License
rate of 98.3 per cent. [08] Plate Classification Method Based on Convolutional
Neural Network” have proposed a novel license plate
Dinesh Bhardwaj et. al [2014] proposed “Comparison classification scheme that is based on convolutional
of ML Algorithms for Identification of Automated neural network. This work uses the CNN seven layers
Number Plate Recognition” has proposed the new that are four convolutional layers which are followed
methodology for number plate recognition in which by max-pooling layers respectively. And then they
there is six imperative parts. In this, the comparative use the fully connect layers that helps in
analysis is carried out among the classifiers based on classification with feature extraction. At last, soft-
factors such as average accuracy, precision, recall, F- max function is performed in soft max layer. The
measure. Here classifiers used are the Random forest faster-CNN is being carried out for the localization of
classifiers, Rep tree, IBK, K-star. Character resizing is the vehicle plate. Here, the 100% classification
done using universal discourser. The feature accuracy is achieved on the training set. It is found

Volume 6, Issue 3, May-June-2020 | http://ijsrcseit.com

960
 Shally Gupta et al Int J Sci Res CSE & IT, May-June-2020; 6 (3) : 955-966

that the proposed research offers the best 98.79 per For the character segmentation the canny edge
cent accuracy score. [11] detection technique is used. Further recognizing
characters is based upon tenser flow framework that
Zied Selmi et. al [2017] proposed “Deep Learning uses second CNN model with 37 classes. The dataset
System for Automatic License Plate Detection and used here are collected from the Caltech dataset and
Recognition” have discussed the automatic license the AOLP dataset. [12]
plate detection and recognition based on deep
learning approach. In this study, using the first CNN This paper presents a survey of techniques focused on
model for LP detection, the author uses the pre- ANPR used for smart Intelligent Transportation
processing steps to identify license plates and non- Network. This study will help the researchers
license plates. And for the classification and identify potential areas of research.
recognition purpose, the second CNN model is used.

IV. COMPARITIVE STUDY

TABLE I. COMPARITIVE STUDY OF PREVIOUS WORK

Authors / researchers Work Method used Result

Rasheed et al. Automated number Hough lines using The result is capable to
(2012) plate recognition using hough transformation discover and acknowledge
hough lines and and template matching the license plates of
template matching Islamabad. It gives out the
total accuracy of 89.70
percent with over all
recognition time to be as 0.9
second for one particular
image.
Sasi et al. Automatic car number Modified Ant colony Result shows the
(2017) plate recognition optimization (ACO) for segmentation and extracting
plate location, image of plates in proposed
hierarchical grouped algorithms gives a precision
structure that is based rate of 94 percent.
on inductive RULES-3 Additionally the combined
and Support Vector approach offers improved
Machine (SVM) results for character
approach recognition.
Subhadhira et al. License plate Extreme learning The result provides the
(2014) recognition application machines (ELM) for acknowledgement of Thai
using extreme learning recognition and number plates. It provides
machines Histogram of oriented an accuracy rate of province
gradients (HOG) for part with 95.05 percent and
feature extraction and to equate license plate with

Volume 6, Issue 3, May-June-2020 | http://ijsrcseit.com

961
 Shally Gupta et al Int J Sci Res CSE & IT, May-June-2020; 6 (3) : 955-966

pre-processing province both results as

89.05 percent efficient.
Shima Extraction of number Morphological image The proposed device is made
(2016) plate images based on processing, pre-trained for the extraction of license
image category convolutional neural plate which shows the
classification using deep network (CNN) Alex- output of 126 rear sample
learning Net for extracting images with 89.7 percent
characteristics, Support success rate.
vector machine (SVM)
as classifier
Bhardwaj and Gujral Automated Number Kstar machine learning To increase the speed of
(2014) Plate Recognition algorithm, sobel and number plate recognition,
System Using Machine vertical segmentation, this method is proposed
learning algorithms with the aid of five steps
(Kstar) based on Indian plate
character analysis.
Liu et al. A hierarchical license Hierarchical approach Result is shown on the base
(2017) plate recognition combining supervised of images taken from 1530
system using supervised K-means and support samples of Taiwan license
K-means and Support vector machine plate for identifying
Vector Machine characters. The accuracy
rate of the device is
observed as 98.89 percent
for unfocused and inclined
images. The system was
observed with high
accuracy rate and less
computations with
decreases complexity.
Li et al. Vehicle License Plate Algorithm based on The combined approach of
(2019) Recognition Combing Maximum stable region MSER and SVM offers
MSER and Support extremum (MSER) and better character recognition
Vector Machine in A Support vector results compared to
Complex Environment machine, Kernel conventional methods
method without the use of position
of license plate and
segmentation of character.
The system proposed is
more concise and effective,
which gives 90 per cent
accuracy.

Volume 6, Issue 3, May-June-2020 | http://ijsrcseit.com

962
 Shally Gupta et al Int J Sci Res CSE & IT, May-June-2020; 6 (3) : 955-966

Guang License plate character Mexico hat wavelet LS- Result shows the accurate
(2011) recognition based on SVM (least square and quick recognition of
wavelet kernel LS-SVM SVM) conveyance license plate.
Experiment was performed
using 500 image samples for
training and testing data. In
addition, the findings of the
proposed study were
compared with the neural
network BP and the LS-
SVM RBF. Among which it
was observed that Mexico
hat wavelet LS-SVM depicts
high success rate as 98.3
percent.
Bhardwaj and Kaur Comparison of ML Kstar ML algorithm, The result is obtained to
(2014) algorithms for WEKA tool, IBK, compare the several
identification of Random Forest, REP algorithms of machine
Automated Number Tree, MLP learning for recognizing the
Plate Recognition license plate. Among all the
algorithms compared it was
observed that kstar depicts
the highest accuracy on the
data sample images that is
99 percent.
Satsangi et al. License Plate Viola Jones ML Recognition of characters is
(2018) Recognition: A algorithm performed on 510 images by
Comparative Study on using Viola Jones ML
Thresholding, OCR and algorithm. In the result the
Machine Learning output comparison was
Approaches carried out with the other
traditional techniques like
threshold and OCR
technology. Among all three
technologies the viola jones
gives out the more efficient
result for recognition that is
80 percent.
Ni et al. A Proposed License Convolutional Neural The result is obtained by
(2018) Plate Classification Network, Faster-CNN using CNN model for the
Method Based on classification of number

Volume 6, Issue 3, May-June-2020 | http://ijsrcseit.com

963
 Shally Gupta et al Int J Sci Res CSE & IT, May-June-2020; 6 (3) : 955-966

Convolutional Neural plate. To avoid the bad

Network quality picture that affect
the successful recognizing of
characters from data sample,
this improved classification
process is added. It was
found that the designed
program was more accurate
as it provided 98.79 percent
accuracy rate.
Selmi et al. Deep learning system Second CNN model, Result demonstrates vehicle
(2017) for automatic license Canny Edge Detection number plate identification
plate detection and Technique, Tenserflow by using deep learning
recognition framework architecture. Tenserflow
framework was opted for
classification and CNN for
recognition. The caltech and
AOLP dataset was used on
2400 dataset images.
Measured with previous
work, the performance was
further measured, and it was
found that the proposed
work offers the best
accuracy rate of recognition.

Issues and Challenges in using ANPR System acquire more vehicle information and to increase the
knowledge for interested vehicles in field of traffic
ANPR system has gained the attention of several management. The recognition of vehicles can be
people in the field of intelligent transport system in done by using various methods but the more efficient
the last few years. The involvement of human way in the traffic management is to utilize ANPR
activities has decreased to a greater extent, with the systems by performing image processing protocols for
help of automatic license plate recognition. It can finding the vehicle information. Recognition systems
track traffic automatically by detecting and are mainly used for monitor traffic, collecting
recognizing the plates on the vehicle. The highly electronic tolls and law enforcement. Other
significant factor in the metropolitan city is the necessary area of interest and scenario for application
management of traffic and finding out the vehicle are repossession of stolen vehicles, control on parking
owner. As observed, the increase in population lots, criminal investigations etc. An open problem is
indicates the growth in the vehicle is in many folds. the proper detection that may be due to distinct
So, recognizing license plate will also be needed to vehicle plate and bad environment conditions [4].

Volume 6, Issue 3, May-June-2020 | http://ijsrcseit.com

964
 Shally Gupta et al Int J Sci Res CSE & IT, May-June-2020; 6 (3) : 955-966

The images may be taken under bad lightening tilted or side view images, moving distance images,
conditions and also camera resolutions may not be numbering scheme, and number plate style
very high. Some other factors that affect the ANPR (background) can be further enhanced. Object
are due to inclination problems, plates not clear detection and neural networks are useful for
enough to be read and also many vehicle plates in a detecting side view or tilted images and moving
picture [12]. The other problems observed were that distance images. This work opens new possibilities by
tilted or side view images and moving distance using deep learning algorithms for image
images were not recognized easily. The framework segmentation processes by the aid of You-Only-
needs the implementation of appropriate image Look-Once (YOLO) algorithm, including
processing techniques to enhance the identification convolutional neural networks for character
of vehicles. Creating a necessary segmentation and recognition. It had also been manifested that the
recognition methods for recognizing of vehicle are ANPR are light weight and a robust method. In
considered to be a big request for the mass future recognition systems have the option of using
surveillance of traffic. It should be kept in mind that high resolution cameras with increased number of
the precision accuracy of recognition is higher and frames for high accuracies and for better recognition
computational complexity to be lower in the city results.
traffic observation to monitor traffic. Another issue
includes facts that the images are noise sensitive and VI. REFERENCES
also sometimes the vehicle plate pictures are full of
dirt. There are missing number plates and dingy [1] Rasheed, S., Naeem, A., & Ishaq, O. (2012,
license plates that also cause problem in accurate October). Automated number plate recognition
identification of vehicle. So, these, in effect, avoid using hough lines and template matching. In
attempts at using ANPR techniques in traffic Proceedings of the World Congress on
management areas. Unless object detection methods Engineering and Computer Science (Vol. 1, pp.
and recognition methods are substantially improved, 24-26).
the recognition of conveyance by ANPR [2] Sasi, A., Sharma, S., & Cheeran, A. N. (2017,
implementation would remain a huge problem. March). Automatic car number plate
recognition. In 2017 International Conference
V. CONCLUSION on Innovations in Information, Embedded and
Communication Systems (ICIIECS) (pp. 1-6).
This paper depicts a survey of recognizing vehicle IEEE.
plate number in traffic surveillance. System with [3] Subhadhira, S., Juithonglang, U., Sakulkoo, P.,
powerful image processing technique can easily & Horata, P. (2014, March). License plate
detect vehicles of interest from different angles and recognition application using extreme learning
display information about the owner as output. This machines. In 2014 Third ICT International
is a much more accurate way of monitoring traffic Student Project Conference (ICT-ISPC) (pp.
than a manual approach. And for that, with the aid of 103-106). IEEE.
the new technology, ANPR systems are playing a [4] Shima, Y. (2016, December). Extraction of
major role in developing the intelligent number plate images based on image category
transportation network. Recognition may use the classification using deep learning. In 2016 IEEE
image processing technique combining with neural International Symposium on Robotics and
networks to detect the number plates where the Intelligent Sensors (IRIS) (pp. 19-26). IEEE.

Volume 6, Issue 3, May-June-2020 | http://ijsrcseit.com

965
 Shally Gupta et al Int J Sci Res CSE & IT, May-June-2020; 6 (3) : 955-966

[5] Bhardwaj, E. D., & Gujral, E. S. (2014, June). Conference on Network Infrastructure and
Automated Number Plate Recognition System Digital Content (IC-NIDC) (pp. 344-347). IEEE.
Using Machine learning algorithms (Kstar). In [12] Selmi, Z., Halima, M. B., & Alimi, A. M. (2017,
International Journal of Enhanced Research in November). Deep learning system for automatic
Management and Computer Applications, license plate detection and recognition. In 2017
ISSN:2319-7471 Vol.3 Issue 6, June- 14th IAPR International Conference on
2014,pp:(42-47), Impact Factor:1.147, Available Document Analysis and Recognition (ICDAR)
online at : www.erpublications.com . (Vol. 1, pp. 1132-1138). IEEE.
[6] Liu, W. C., & Lin, C. H. (2017, May). A
hierarchical license plate recognition system
using supervised K-means and Support Vector Cite this article as :
Machine. In 2017 International Conference on
Applied System Innovation (ICASI) (pp. 1622- Shally Gupta, Rajesh Shyam Singh, H.L. Mandoria,
1625). IEEE. "A Review Paper on Automatic Number Plate
[7] Li, Y., Niu, D., Chen, X., Li, T., Li, Q., & Xue, Y. Recognition System", International Journal of
(2019, July). Vehicle License Plate Recognition Scientific Research in Computer Science, Engineering
Combing MSER and Support Vector Machine and Information Technology (IJSRCSEIT), ISSN :
in A Complex Environment. In 2019 Chinese 2456-3307, Volume 6, Issue 3, pp.955-966, May-June-
Control Conference (CCC) (pp. 7045-7050). 2020. Available at
IEEE. doi : https://doi.org/10.32628/CSEIT2063208
[8] Yang, G. (2011, March). License plate character Journal URL : http://ijsrcseit.com/CSEIT2063208
recognition based on wavelet kernel LS-SVM.
In 2011 3rd International Conference on
Computer Research and Development (Vol. 2,
pp. 222-226). IEEE.
[9] Bhardwaj, D., & Kaur, H. (2014, October).
Comparison of ML algorithms for identification
of Automated Number Plate Recognition. In
Proceedings of 3rd International Conference on
Reliability, Infocom Technologies and
Optimization (pp. 1-6). IEEE.
[10] Satsangi, M., Yadav, M., & Sudhish, P. S. (2018,
October). License Plate Recognition: A
Comparative Study on Thresholding, OCR and
Machine Learning Approaches. In 2018
International Conference on Bioinformatics
and Systems Biology (BSB) (pp. 1-6). IEEE.
[11] Ni, K., Fu, M., Huang, Z., & Sun, S. (2018,
August). A Proposed License Plate
Classification Method Based on Convolutional
Neural Network. In 2018 International

Volume 6, Issue 3, May-June-2020 | http://ijsrcseit.com

966