Automatic Numberplate Recognisation 1

CHAPTER 1
INTRODUCTION

1.1 Background

Automated Number Plate Recognition (ANPR) is also known as Automated License Plate
Recognition (ALPR). Automatic Number Plate Recognition or ANPR is a technology that
uses pattern recognition to 'read' vehicle number plates. In simple terms ANPR cameras
'photograph' the number plates of the vehicles that pass them. This 'photograph' is then fed in
a computer system to find out details about the driver and owner of the vehicle and details
about the vehicle itself. ANPR consists of cameras linked to a computer. As a vehicle passes,
ANPR 'reads' Vehicle Registration Marks - more commonly known as number plates - from
digital images, captured through cameras located either in a mobile unit, in-built in traffic
vehicles or via Closed Circuit Television (CCTV). The digital image is converted into data,
which is processed through the ANPR system. Here we proposed a method mainly based on
edge detection, OCR operation and Finding Rectangles in a Vehicle Image.

Generally, the ANPR technology can be bought in two modalities:

 The ANPR engine
 The ANPR equipment (Hardware + recognition engine)

The ANPR engine can recognize the number plate directly from the images stored in a hard
disk. Software of this type allows for taking efficient use of images that have been received
from other systems like CCTV or cameras. Scene analysis running on the automatic licence
plate recognition platform becomes very demanding, which introduces many new difficulties
to object detection tasks and algorithms. These challenges mainly include, how to manage the
various changes commonly encountered in the visual aspects of objects in attaining images
(for example, light, vision, small size, and ratio)
 Automatic Numberplate Recognisation 2

1.2 Literature Survey

The list of papers cited relevant to the work is given below

Author Work Method Used Result

Sasi et al. Automatic car Modified Ant colony Result shows the
(2017) number plate optimization (ACO) for segmentation and
recognition plate location, hierarchical extracting image of plates
grouped structure that is in proposed algorithms
based on inductive RULES- gives a precision rate of 94
3 and Support Vector percent. Additionally, the
Machine (SVM) approach combined approach offers
improved results for
character recognition.

Li et al. (2019) Vehicle License Algorithm based on The combined approach of

Plate Recognition Maximum stable region MSER and SVM offers
Combing MSER and extremum (MSER) and better character
Support Vector Support vector machine, recognition results
Machine in A Kernel method compared to conventional
Complex methods without the use of
Environment position of license plate
and
segmentation of character.
The system proposed is
more concise and
effective, which gives 90
per cent accuracy.

Ni et al. (2018) A Proposed License Convolutional Neural The result is obtained by

Plate Classification Network, Faster-CNN using CNN model for the
Method Based on classification of number
Convolutional Neural plate. To avoid the bad
Network quality picture that affect
the successful recognizing

2
 Automatic Numberplate Recognisation 3

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.

1.3 Motivation
From the literature survey, it is observed that very less work cited with respect to
identification of numbers from the boards and hoardings, vehicle number plates, and
advertisement banners using deep learning techniques. This forms the basis for the proposed
project the problem statement is framed as “To implement a number recognition system from
vehicle number plates using advanced deep learning architectures”.

1.4 Objectives

i. To construct the image database of licence plates

ii. To extract frames from videos.
iii. To implement suitable segmentation technique to extract number plate from the
images.
iv. To deploy DCNN architecture for number and character recognition
v. To evaluate the performance of different DCNNs.

1.5 Problem Statement

“To implement a number recognition system from vehicle number plates using advanced
deep learning architectures”.

1.6 Scope

3
 Automatic Numberplate Recognisation 4

 Automatic number plate recognition cameras are used to measure the average vehicle
speed over longer distances.
 Automatic number plate recognition cameras are used for Traffic management
systems.
 ANPR camera solutions automatically recognize customers based on their license
plate and provide them the complete information about the items that they ordered the
last time they used the service.
 Automatic license plate recognition camera solutions are used to recognize the guest
vehicles in order to assist visitor management systems.

1.7 Limitations
 ANPR does not work if vehicles are travelling over 30mph.
 Plates not big enough to read (incorrect camera, lens or location).
 Poor on-site Traffic management resulting in missed plates/vehicles.
 Excessive skew angles causing recognition issues.

CHAPTER 2
2.1 Proposed Methodology
The block diagram of the licence plate recognition methodology is given below

4
 Automatic Numberplate Recognisation 5

Figure 2.1: Block diagram of the licence plate recognition methodology.

Here we cast a framework which detects and recognizes Licence Plates in a given image. Our
approach is designed to focus on finding and reading LPs in tough environments. This type of
method can be divided into three main parts:

• Licence Plates Detection

• Character Segmentation

• Character Recognition

2.1.1 YOLO3

YOLOv3 is the latest variant of a popular object detection algorithm YOLO-You Only Look
Once. The published model recognizes 80 different objects in images and videos.

The Object Detection using Yolov3, the trained our model from a combination of exclusive
manually annotated data and a pre-existing weights file. The algorithm detects the bounding
boxes of each license plate in each frame received from the camera. YOLO divides the input
into n*n grids first. Once the threshold confidence is achieved in the detection then the frame
is saved and proceed to the next phase.

After Licence Plate detection, first introduce some padding (1-2 pixels) in order to improve
prediction because some characters might now be well segmented. In order to train the
networks, segmented characters with labels are passes Units as input. For recognition, the
CNN algorithm is used, which is designed for text recognition. Given the Licence Plate,
containing the characters, features extracted using CNN is transformed into feature vectors
and then used as an input for the CNN layer, helping sequence layer problem and predicting a
label distribution.

5
 Automatic Numberplate Recognisation 6

2.2 Resources required:

2.2.1 Hardware Specifications
The hardware specifications are listed as below.

Processor: i5

RAM:8GB

Harddisk:1TB or above

2.2.2 Software Specifications

The software specifications are listed as below.

Python 3.6+

Sklearn library for Machine Learning

OpenCV for image processing

2.3 Applications:
 Parking: The number plate is used to automatically enter pre-paid members and
calculate parking fee for non-members (by comparing entry and exit timings).
 Tolling: The car number is used to calculate the travel fee in a toll-road, or used to
double-check the ticket.

2.4 Budget
13000

REFERENCES:

6
 Automatic Numberplate Recognisation 7

[1] Sasi, A., Sharma, S., & Cheeran, A. N. (2017, March). Automatic car number plate
recognition. In 2017 International Conference on Innovations in Information,
Embedded and Communication Systems (ICIIECS) (pp. 1-6). IEEE.
[2] Li, Y., Niu, D., Chen, X., Li, T., Li, Q., & Xue, Y. (2019, July). Vehicle License Plate
Recognition Combing MSER and Support Vector Machine in A Complex
Environment. In 2019 Chinese Control Conference (CCC) (pp. 7045-7050). IEEE.
[3] Ni, K., Fu, M., Huang, Z., & Sun, S. (2018, August). A Proposed License Plate
Classification Method Based on Convolutional Neural Network. In 2018 International
[4] 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.
[5] Selmi, Z., Halima, M. B., & Alimi, A. M. (2017, November). Deep learning system
for automatic license plate detection and recognition. In 2017 14th IAPR International
Conference on Document Analysis and Recognition (ICDAR) (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 Machine. In 2017 International
Conference on Applied System Innovation (ICASI) (pp. 1622- 1625). IEEE.