International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 04 Issue: 07 | July -2017 www.irjet.net p-ISSN: 2395-0072

Automatic Number Plate Recognition and IoT Based Vehicle Tracking

Avadhut S Joshi1, Digambar A Kulkarni2

1,2 Dept.
Of CSE, Gogte Institute of Technology, Belagavi, Karnataka, India.
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - License plate recognition (LPR) plays a significant Automatic Number Plate Recognition (ANPR) falls under this
role throughout this busy world, owing to the rise in vehicles. category which is a method to recognize the characters on
Segmentation plays an important art, to avoid problems like the images which can be captured using the traffic cameras
unwanted illumination, tilt that degrades the segmentation or CCTVs. Vehicle registration Plate location, normalize and
which in turn affects the recognition accuracy algorithms are improvise the amount of plate, then the identification of
developed for this work. The work here presents a strong character are the traditional steps followed in all Image
technique for localization, segmentation and recognition of processing techniques.
the characters within the located plate. Images from still
cameras or videos are obtained and regenerated in to Plenty of systems that play the role of server are used in
grayscale images. Hough lines are determined using Hough order to process the data received from the roads. Among
transform and therefore the segmentation of grey scale image these many systems just do the work of transferring the
generated by finding edges for smoothing image is employed information to the systems which play a role of remote
to cut back the quantity of connected part and then connected servers.
part is calculated. Finally, single character within the
registration code is detected. It is necessary for every In this literature work, the implemented system uses
individual to track the vehicle. The proposed system will also template matching for number plate recognition of vehicle
be used for the tracking the vehicle from remote location. The entry from a medium resolution camera. And also this work
theft vehicles, vehicles violating traffic rules can be tracked helps to track the vehicles location of the vehicles on the web
easily by making use of IOT based vehicle tracking. or android phone using the GPS and mobile data.

Key Words: Number Plate, IoT, Segmentation, 1.1 EXISTING SYSTEM AND ITS LIMITATIONS
Recognition, Tracking, Optical Character Recognition,
Support Vector Machine Identification of the characters and various algorithms for
registration plate identification contribute positively in
1.INTRODUCTION registration plate image analysis. Therefore, they are the
heart of any ANPR system. The ANPR system encompasses a
Now a days, recognition or identifying the information camera, software needed for image processing, a computer
present on the moving objects is turning of great importance. system, and frame grabber for analysis and recognition.
The increasing human dwelling and the growth of industry Number Plate Recognition has gained a lot of importance in
arena in the current scenario has made vehicle a mandatory the last few years. Considerable amount of research work
need which in turn leads in serious traffic issues. The resting has been carried out to recognize the vehicle type like car,
area for vehicles i.e. parking lots face overflow of vehicles truck. The Support Vector Machine is used to identify the
due to the above mentioned problem. vehicle model. They demonstrated the numerical results on
more than 50 data set pictures.
One of the methods used for this purpose is, background
subtraction or foreground segmentation. Background There are adequate number of applications and software
subtraction or foregrounds segmentations process of available in the market to identify or recognize the
separation of objects in video or image and is a ubiquitous registration number plates. These systems serve the purpose
procedure in almost all the computer vision applications. with few limitations. To cite a few: these applications
With growing trends and enhancements in recent consume ample amount of time for the recognition also they
communications technologies has necessitated the need for require image with high quality. Even though, these
Intelligent Surveillance Systems (ISS) technologies. This applications recognize the character on the number plate,
need facilitates obtaining information about the traffic, they fail to achieve the accuracy. The existing ANPR systems
accidents and safety management. Operational efficiency can recognize the number plate alone but lag in tracking the
be improvised by vision based information. Separation of vehicles location.
dynamic objects in picture sequences, is an emerging and a
terribly beginning in this respect. There exist standard Indian registration plates, but they are
seldom used. As a result, wide variety of number can be
found in terms of character size, type of the font and location

2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 1503
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 07 | July -2017 www.irjet.net p-ISSN: 2395-0072

of the registration plate and unnecessary characters are

present on the number plate in many cases,
The following figure shows the Indian Standard Number
Plate format which has first two characters as State Code,
next two numbers represents District code following by the
two characters for series code and finally a four-digit
Registration number.

KA22 CS 8055

FIGURE 1 Indian Standard Number Plate Format

1.2 PROPOSED SYSTEM

In present days, there is a damn need of smart

transportation system due to growing world & importance of
the time in day to day life. Hence we provide system that
provides the Smart Transportation System information of
vehicle tracking and monitoring. In this proposed system, we
recognize the vehicle number plate through the cameras,
CCTVs which will be converted from RGB image to Optical
Character Recognition (OCR). This system also provides the
feature for tracking the vehicle using its GPS location easily.
The location of the vehicle can be traced and updated every
ten seconds to track the vehicle continuously using Map
along with data storage. Here, every vehicle need to carry the
GPS. The GPS continuously sends the vehicle location to the FIGURE 2 System Architecture
server for the registered vehicle. By integrating both IOT
technology and Image processing, we are going to design a There are several processing stages takes place in order to
whole Number Plate Recognition and vehicle tracking recognize the number plate as mentioned below:
system. Whenever we want to track any GPS enabled vehicle,
we can track it and also identify the registration plate in OCR Localization: Localizing registration plate is performed to
form easily. eliminate the background which is unwanted details
meaning that, we concentrate only on the number plate and
2. ARCHITECTURE OF THE SYSTEM rest of the area are ignored and hence the focusing on to the
required details on the registration plate. To detect the
System architecture describes the diagrammatic registration plate, we apply a top-flat filter to the complete
representation that has the building blocks of the system image begin by a multistage region search. One more
called components. During the implementation phase, these approach has been proposed to consider the vertical edges,
conceptual components will be transformed into working for extracting the registration plate using sable operator. An
system components. Input to this literature is the image edge detection and Hough transforms techniques are used to
captured from the camera with a medium quality or high detect the vertical and horizontal edges, by using the
quality resolution. As shown in Figure 4.1, which illustrates registration plate in a rectangular shape. Image captured
the system flow for registration plate recognition by from the camera from 8-10 meters away can be processed
performing locating the image, segmenting the image, through the registration plate extraction with providing its
detection of the image from the registration plate. output to segmentation.

Extraction of Characters: Extraction of characters is

performed by separating the individual character area from
the localized registration plate. This image contains the noise
along with the actual image which is a tedious job, hence it is
required to remove or eliminate the noise from the
registration plate. The next process is performed called a
digit filtering to remove the frame lines from the registration
plate. This system solves the problem of extracting the
registration plate on a higher level, by integrating the

2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 1504
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 07 | July -2017 www.irjet.net p-ISSN: 2395-0072

outcome of different threshold methods, also implementing Median Filtering: To eliminate the unwanted noisy regions,
the feedback from character recognition and character median filter is used. Median filtering is a technic that passes
phases has been proposed. the image between 3X3 matrix. The dimension can be
modified according to level of noise for this matrix. The
Recognition of Character: This technique plays a vital role progress follows as mentioned below.
in this application to recognize the character extracted from
the segmentation. The actual conversion of Number plate The matrix of 3X3 is chosen for one pixel.
image into Optical Characters is performed in this phase and The pixel around the neighborhood are assigned.
the clear identification of number plate is performed. For The process of sorting is developed between nine
character recognition, we use the data set of individual pixels from small to big.
character image of size 24X42. The extracted character will Median element can be termed for the fifth element.
match with the similar data set and we get the proper optical The implementation of these procedures to all
character from the extracted image. For the character pixels in the picture.
recognition, every character four data sets are collected of
different variety. The test data set are used as alphanumeric Character Segmentation: MATLAB Provides the function
such as characters A-Z only capital letters and the numbers called Regionprops that identifies number plate region are
from 0-9. segmented to have the bounding for each pixel. The smallest
bounding returned by Regionprops which contains a
Tracking: Tracking is an important concept in this character. This technic is favored to obtain the registration
literature. The theft off vehicles, vehicles violating the traffic plate The segmentation can be done for database character
rules can be tracked using IOT based tracking scheme. To matching with the training data sets.
track the vehicle, the vehicle must be GPS enabled. The
vehicle is tracked for every ten seconds and the exact OCR using Template Matching: The most commonly used
location can be obtained with the help of latitude and technique for image comparison is Template Matching. In
longitude provided by GPS. To work with the GPS, the this technique, we are focusing to get the position of a sub-
android application is developed by using which the owner image called a template, in the given image. The training data
or driver will register the system with valid information like set matching describes the equality in a given training set.
Owner/Driver name, Vehicle registration number, Contact The process of comparing is done by pixel by pixel of the
number and type of the vehicle etc. picture and the find the match for each possible
displacement. It enables the matching of datasets character
3. METHODOLOGY matching with training data. We have used almost all
possible training data sets to match recognized inputs.
The implementation is started by capturing the number Training data set are created to match each character,
plate of the vehicle. The edge detection and gray scale filter number (from A-Z and 0-9) using method 'Regular' font
is applied initially as a preprocessing for selected image to style. Figure 5.1 shows the training data set for some
isolate the number plate region which is a smaller part from alphabets and numeric characters.
extracted image. The gray scale picture is obtained by
quantizing from neutral gray starting from the black as a
weakest intensity to white as a highly strongest intensity.
Now it is required to binaries the by assigning the pixel
values of 0 for black shade and 1 for white shade.

Furthermore, the registration plate is identified by observing

the quick change in the contrast. Rest of the area in the
images made filtered. The actual registration plate location is
located by matching the width and height. The contrast
extension and median filtering techniques enhances the gray
level of registration plate image.

Contrast Extension: The image contrast can be extended by

means of the process of enhancing the contrast. equalization
of histogram of that particular picture. In other words, the
image can be made sharpen by applying the extension of
contrast. In order to improve poor picture contrast view,
histogram equalization is a well-known technic.

FIGURE 3: Alphanumeric Characters as Data sets

2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 1505
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 07 | July -2017 www.irjet.net p-ISSN: 2395-0072

The process of ANPR starts with identifying a registration be a process of finding horizontal boundaries between
plate of the vehicle. It involves the algorithms used which are characters [1].
able to identify rectangular area of registration plate from an
original picture. The identification and recognition process Another method is implemented here called as aggregation
takes place in four phases mainly. of obtained segments. This phase of a number plate having
characters not desirable parts like, hyphen, dashes and dots
Preprocessing of Image stretches further as redundant space on the edges of
Localizing Registration Plate character. It is important to remove these parts and obtain
Segmentation of Characters only the registration plate characters. Since the segmented
Recognition of Actual number plate. plate is deskewed, we could segment it by identifying spaces
in its horizontal projection.
Preprocessing the Image
Recognition
The original picture contains large variety of colors and
hence, in order to enhance the standard, the image is The segmented numbers are compared with data sets from
processed initially and prepared for further phase. The the database. Once an Image is obtained, segmentation is
original image contains many colors so this phase converts done is Grayscale. Before making the model ready, for each
the RGB to gray scale using equation. of the characters for more use, we'd like make some
processing on the images. The subsequent operations that
Gray=0.299*Red+0.587*Green+0.114*Blue are performed like Binarization and Inversion of the
characters. Realize the connected component that represents
In this phase, the median filter filters the grayscale image, in the optical character and realize the smallest rectangular
order to remove the noisy section of the image, but that region.
affects slight tedious because of the sharpness of picture. The
Median filter is used for nonlinear filter which swaps for The image normalization is done with size 24X42. Here we
each and every pixel with a efficiency obtained by calculating store the intensity values below algorithmic rule mentioned
the median filter of different values of pixels. Picture can be for each and every characters. Compute the values for each
sub-divided into multiple groups by of the characters. We need to compute template matching
score of the characters separated from the training data. The
Total number of groups = Height / Candidate training data set contain the large number of data to
compared with the localized plate. Each localized plate is
compared with the SVM i.e. Support Vector Machine
Localizing Registration Plate classifier which exactly matches or similar match occur. So
the template matching is performed by comparing the only
To detect the exact registration plate portion from the given required comparisons.
complete image, this method is implemented. The major goal
of this system is to locate the exact registration plate region Tracking
from the complete picture from the CCTV, camera and video.
The image quality plays important part. Thus localizing The tracking is implemented based on the GPS provided
technic used for increasing the standard. The numbers and along with the vehicle. The owner/driver need to register
letters on a registration plate region must be placed in a the vehicle and other information from android application.
same row. This gives the promotion for identifying changes This is one-time process of registering the vehicle. The GPS
within horizontal intensity because the rows will contain the of the vehicle is always on and hence wherever the vehicle
number plate are expected to show variations. moves, its appropriate location is update and stored on the
web server i.e. its longitude and latitude. The vehicle can be
Before this, the application of Hough transform application is tracked at any point of time by typing the vehicle
given to edge detector for perfect edge detection. These registration number in the ANPR Interface. When a vehicle
vertical and horizontal edges facilitate to characterize the number is typed, the registered vehicles GPS responses with
angle of fundamental. the latitude and longitude to the web server and the
location/map for that particular vehicle is displayed
Segmentation separately on the web browser. The browser updates the
location for vehicle for every ten seconds to keep updated
The further step is segmentation of the registration plate. Its information about the vehicle.
totally clear that this is an important technic. If this process
fails, then the characters will be improperly recognized. If we
try to assume only one-row plates, the segmentation could

2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 1506
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 07 | July -2017 www.irjet.net p-ISSN: 2395-0072

4. CONCLUSIONS [4] H.J. Lee, S.Y. Chen, and S.Z. Wang, Extraction and
recognition of license plates of motorcycles and vehicles on
This work developed completely unique methodology for the highways,, in Proc. ICPR, pp. 356-359, 2004.
recognition of vehicle registration plate. The proposed [5] Hinde Anoual, Sanaa El Fkihi, AbdeilahJilbab and
method uses horizontal projection profile and Hough DrissAboutajdine, A Novel Texture-Based Algorithm for
transform. Localizing Vehicle License Plates,, in Journal of Theoretical
and Applied Information Technology 15th November 2012.
Firstly, the extraction process of number plate from a given Vol. 45 No.1.
input image is located, then performed segmented the plate [6] A. Broumandnia and M. Fathy,Application of pattern
characters independently and applied template matching by recognition for Farsi license plate recognition,, presented at
using template data set for recognition of plate characters. the ICGST Int. Conf Graphics, Vision and Image Processing
Further, it is assured to be approximately 96% for the (GVIP),[Online]. Available:
extraction of plate region, 95% for the separation optical http://www.icgst.com/gvip/v2/P1150439001.pdf, Dec.
characters recognition unit accurate, giving the overall 2005.
system performance better than the existing system. The [7] T. D. Duan, T. L. Hong Du, T. V. Phuoc and N. V. Hoang,
recognition time is between 5 to 15 seconds which is Building an automatic vehicle license plate recognition
relatively shorter than current time i.e. more than 30 sec. system,, Iin Proc. Int. Conf. Comput.Sci. RIVF, pp. 59-63,
Also, the vehicle location can be accurately tracked for GPS 2005.
enabled vehicles and the location can be refreshed for every [8] C.T. Hsieh, Y.S Juan, and K.M. Hung, Multiple license plate
ten seconds to update the vehicle information. detection for complex background,, in Proc. Int. Conf. AINA,
vol. 2, pp. 389-392, 2005.
FUTURE WORK [9] D.S. Kim and S.I. Chien, Automatic car license plate
extraction using modified generalized symmetry transform
There are many enhancements are possible with this and image warping,, in Proc. ISIE, pp.2022-2027, 2001.
research in order to add the quality and features. Here are [10] T.Maheswari And P.Sukumar, Error Detection and
some future enhancements possible as below. Correction in SRAM Cell Using DecimalMatrix Code, IOSR
Journal of VLSI and Signal Processing (IOSR-JVSP), Volume 5,
This application can be modeled for international Issue 1, Ver. II (Jan - Feb. 2015), PP 09-14
vehicle registration plates.
The literature can be modified to capture the number
plates which are not in the Indian standard format i.e.
number plate contains two lines.
Further, the regional language data set can be added
to the database and the same will be recognized
accurately.
Some number plate contains the unwanted data
which can be recognized by redesigning the
application.
The detected numbers can inform the user the RTO
location, region, district etc.

REFERENCES

[1] Priyanka Prabhakar, AnupamaP, Resmi S R, - Automatic

Vehicle Number Plate Detection and
Recognition.International Conference on Control,
Instrumentation, Communication and Computational
Technologies (ICCICCT) -1-4799-4190-2/14/$31.00 2014
[2] Abbas M. Al-Ghaili, SyamsiahMashohor, Abdul Rahman
Ramli, and AlyaniIsmail, Vertical-Edge-Based Car License-
Plate Detection Method., IEEE Transactions on Vehicular
Technology vol. 62, no. 1, Jan 2013.
[3] B. Hongliang and L. Changping, A hybrid license plate
extraction method based on edge statistics and
morphology,, IEEE Proc. ICPR, pp. 831-834, 2004.

2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 1507