faculty: aces msc: full time module: dissertation …hsnemer.weebly.com › uploads › 7 › 6 ›...

S h e f f i e l d H a l l a m U n i v e r s i t y P a g e | 1

Name: Harish Kumar Reddy MedipallyStudent ID: 18039734Submission date: July 2nd 2010Supervisor: Dr. Hussein Abdul-Rahman

L i c e n c e P l a t e L o c a l i z a t i o n u s i n g e d g e d e t e c t i o n .

FACULTY: ACESMSC: FULL TIMEMODULE: DISSERTATIONTOPIC: licence plate localization and recognition algorithm using edge detection technique.


Abstract:

Licence plate localization and recognition (LPLR) uses image processing and character

recognition technology in order to identify the licence number plates of the vehicles

automatically. LPLR system is a kind of an intelligent transport system and is also of

considerable interest because of its good applications in the traffic monitoring system,

electronic toll collection, surveillance devices and safety supervision systems. These

applications demand high reliability of LPLR system. The system consists of main modules

Localizing the licence plate

Detecting the licence plate

Extracting the number plate

Recognising the numbers and letters in the plate

The purpose of this report is to investigate and compare between different plate localization

algorithms, and to extract and isolate the licence plate from the rest of the image. The

proposed system has been implemented using MATLAB.

Towards the end we also have discussed the various different approaches implemented and

further improvements to the proposed system to improve the overall accuracy and efficiency

of the system.



Acknowledgement:

I would have never succeeded in completing my work without the encouragement,

cooperation and help provided to me by my supervisor.

With deep sense of gratitude I express my sincere Thanks to my esteemed and worthy

supervisor, Hussein Abdul-Rahman for his valuable guidance in carrying out this work

under his effective supervision, encouragement, enlightenment and cooperation with whose

assurance and kind advice brought me closer to the realization of this project.

I shall be failing in my duties if I don’t express my deep sense of gratitude toward Dr.

Abbass Hashim (Research Fellow and associate lecturer), for his great suggestions,

evaluation and encouragement which helped me in improving my project.

I would like to thank you for your time, patience, generous assistance and sincere interest in

my work.



Organization of Thesis:

Chapter 1: This chapter gives a brief introduction about the problem, previous approaches,

aims and objectives and methodology.

Chapter 2: The first chapter briefly gives the literature review and the work that has been

done previously.

Chapter 3: This chapter gives a brief introduction to the system elements, its applications,

working and the structure of the proposed system.

Chapter 4: This chapter shows the detailed description of the analysis and the available

processing tools in the application software on which the work is being done.

Chapter 5: This chapter gives the problem definition and the proposed system solution and

about the implementation of various tools of application software for simulation and the

testing part of the thesis.

Chapter 6: This chapter finally discusses about the conclusion to the thesis with future

scope.



Table of contents:

DeclarationAbstract 2Acknowledgement 3Organization of Thesis 4

1.chapter: Introduction 1.1Introduction 7 1.2Aims and objectives 8 1.3Methodology 8 1.4 Outline 9

2.chapter: Literature Review 2.1Introduction 10 2.2LPR and OCR 11 2.3Image acquisition (capturing) 11 2.4Licence Number Plate Extraction 11 2.5Segmentation 12 2.6 Character recognition 13 2.7Summary 14

3.chapter: Software Development& image processing in Matlab. 3.1 Introduction 15 3.2Digital image 15 3.3Images in Matlab 15 3.4 Working formats in Matlab 16 3.4.1 Binary image 16 3.4.2 Intensity image 16 3.4.3 Indexed image 16 3.4.4 RGB image 16 3.4.5 Multi frame image 163.5 Conversion between different formats 173.6 Reading and writing image files 173.7 Saving variables in Matlab 173.8 Displaying image in Matlab 18

4.chapter: Requirement Analysis:4.1Introduction of LPLR 194.2Applicationsof LPLR system 194.3 Required Elements for system 214.4Colour Models 22 4.4.1RGB colour space 22 4.4.2 HSV and HSI colour spaces 234.5 Optical character recognition 24 4.5.1 Introduction 24 4.5.2 Applications 24 4.5.3 Training characteristics 254.6 Thresholding 26



4.7Algorithms 26 4.8Working of system 29 4.9Structure of the proposed system 30 4.9.1colour extraction method 30 4.9.2edge detection and extraction method 32 4.9.3 Pre processing 33 4.9.4 Plate extraction 33 4.10segmentation 37

5.Chapter:Proposed solution and Experiment results : 5.1Problem definition 39 5.2 proposed solution 39 5.3 Results 39 5.4problem encountered 51 6.chapter: Discussion and Conclusion: 6.1 conclusion 52 6.2 suggestions for future scope 52References: 53



Chapter 1: Introduction

1.1 Introduction:

Licence plate localization and recognition applies image processing and character recognition system to identify vehicles by automatically reading their images. Vehicle licence plate constitute of an identifier of the vehicle participating in the road traffic. This system can be used at every step of the traffic and road monitoring systems. Reading or locating the licence number plate is the main and the first step in determining the identities of parties involved in the traffic incidents. An efficient licence plate localization system may become the core of fully computerised road traffic monitoring systems. Main applications of this system include private parking lot management, traffic managing, automatic traffic ticket issuing, toll collection and security enforcement. Furthermore whenever the data gathered by the system is stored and organised in a database, more complex information-driven tasks can be achieved like vehicle travel time calculations, border control and marketing analysis. This thesis presents the Licence Plate localization system as an application of the computer vision. Computer vision is a process of extracting a high level information from a digital image using a computer.

There are different standards found all over the world for licence plates and there are also variations in each country which are limited but they are slightly different in their styles like in most of the European countries. The figure below shows different types of number plates used in different countries.

Fig.1.1. Sample European license plates from [Kustermann, 2004]

These licence plates are characterised by high contrast between the characters and the uniform backgrounds.



In order to identify a vehicle by reading its licence plate successfully it is necessary to locate the plate in the scene image provided by some acquisition systems like a video or still camera. For example, a currently common1024x768 resolution image contains a total of 786,432 pixels, while the region of interest (in this case a license plate) may account for only 10% of the image area. Also, the input to the following segmentation and recognition stages is simplified, resulting in easier algorithm design and shorter computation times. Due to the current performance of Optical Character Recognition systems described in the next chapter, colourful and textured backgrounds should be eliminated prior classification of alphanumeric symbols.A number of commercial software in this area but they cannot be readily used when vehicle image is provided in different styles and formats. most of these software presume some constrains on the position and the distance from the vehicle to the camera, the complexity and the inclined angles of the captured image.

Image processing techniques like edge-detection, thresholding, re-sampling and filtering have been used to locate the image and isolate the plate and the characters from the image. These techniques alone will not be sufficient to meet the requirements of modern systems. An intelligent license plate localization and recognition system today is required to operate robustly in environments with complicated backgrounds and light intensity variations. To deal with such problems, researchers have proposed various solutions to address these problems.

1.2 Aims & Objectives:

The main Aim & objective of this thesis is to develop a system which is able to localize and extract the licence number plate from the complex scene images without any assumptions like the angle in which the plate has been captured and the distance from the number plate and the camera(unless if it is more far) etc.The work presented here aims at the following aspects

Investigate and implement some plate localisation techniques and, if possible, enhance the existing techniques to overcome some challenges and limitations.

Study the existing licence plate localization and recognition systems Build a system that delivers optimal performance in terms of both speed and accuracy. Suggesting ideas for the future scope of the better plate recognition systems.

1.3Methodology:

The proposed methodology has four stages. In the first stage, it will get input image through high resolution digital camera. In second stage, license plate is extracted using vertical edge matching technique. Then the elements (characters and numbers) in the extracted license plate are segmented by utilizing vertical and horizontal scanning method in the third phase. In the final stage, if time allowed, characters and numbers are recognized by using template matching and fuzzy logic combination.



1.4 Outline:This thesis is organised as follows. A briefing of previous works and literature review is discussed in chapter 2.the proposed model and the steps of the system are explained in the chapter 3. The software development and the techniques involved in the system are discussed in the chapter 4. Simulation and the experimental results are presented in chapter 4 and at last the conclusions and the future work are presented in the chapter 6.



Chapter 2:Literature Review

2.1Introduction :

Licence Plate Localization and Recognition system (LPLR) has become the most important application in the present day transportation system. it is used in many applications like speed control, parking control, road traffic, toll collection, security and entrance admission etc. because of its vital role and various applications this system has received a lot of attention from research communities. Much research has been done on this system in many countries like Korea, China, England and US. A feature of research work in this area is restricted only to a specific region because of lack of standardization between the licence plates of different countries like layout and different styles of number plates).This section given an overview of the techniques employed in improving the efficiency of system and the various research carried out so far in this area. Generally the LPLR algorithm consists of the following steps1) Capturing the vehicle’s image2) Extraction of licence plate from the image3) Extracting the characters from the plate image 4) Finally, Recognizing license plate characters.

Fig:2.1 System development method used till character segmentation.



2.2 LPR and OCR :

Perception’s is the first corporation that launched the commercially available computer vision system for reading the licence plates on moving vehicles automatically. This technology is available since 20 years. Many commercial products usually operate with monochrome acquisition system but they can perform accurately only on some issued plates which have high contrast of colour and clean background and characters. In some situation where there is a little control of environment the approach in target acquisition is to select sensors carefully in such a way to enhance the objects of interest and diminish the irrelevant details like infra-red imaging. New manufacturing techniques are introduced which are more optical recognition friendly. With these techniques the irrelevant details in the background of number plates, when illuminated for machine readability in the near infra-red region become uniform and the plate becomes highly contrasted.

Optical character recognition (OCR) is the electronic or mechanical translation of images of handwritten or typewritten text into machine editable text. Modern commercial OCR systems have been available since 1950’s. Most of the commercial systems may be able to handle different type of character fonts and sizes but they cannot handle text printed characters against shaded or textured backgrounds. Some systems use page segmentation system before OCR but they assume specific layout and they need clean binary input. in some systems if gray scale input is admitted then global thresholding is applied prior to recognition which will fail if there is texture background.

2.3 Image Acquisition :

In general licence plate localization and recognition systems consists of four phases, they are image acquisition, plate localization, plate extraction and character segmentation. Image acquisition is the first step in the LPLR system and there are number of ways to acquire images. Now here in this literature we discuss about different type of image acquisition methods used by different authors. Naito et. al. [13,14,16] developed a sensing system in which he used two CCDS (charge coupled devices) and a prism to split an incident ray into two lights with different intensities. The main feature of the system is to cover wide illumination conditions from twilight to noon under sunshine, and this system is capable of capturing images of fast moving vehicles without any blur in the images. Yan et. al.[20] has chosen a method by using a image acquisition card that converts video signals to digital images based on some hardware-based image pre processing.Comelli et. al.[6] Used a TV camera and a frame grabber card to capture the image for the developed vehicle system. Saldago et. al.[15] used a sensor subsystem having a high resolution CCD camera which is supplemented with a number of new digital operation capabilities.

2.4 Licence number plate extraction:

Licence plate extraction is the second and the most important phase in the LPLR system. The accuracy of the system depends considerably on this phase. This section discusses some of the previous work done during the extraction of the number plate by various authors. Bremananth, chitra and seetharaman et al.[3] proposed weight based density map(WBDM) method which is well suited for application in countries where the licence plate exterior fluctuates extensively. D. Zheng, Y. Zhao, and J. Wang and B.hongliang et al.[4] used edge



statistics and mathematical morphology combination to extract the number plate and high licence plate extraction rate is attained by this method. Kim et. al. [17] has used two Neural Network-based filters and a post processor to combine two images which are filtered in order to locate the licence plates. The two neural networks used are vertical and horizontal filters, which examine small windows of vertical and horizontal cross sections of an image and decide whether each window contains a licence plate. Cross-sections have sufficient information for distinguishing a plate from the background. Hontani et. al. [21] proposed a method for extracting characters without prior knowledge of their position and size in the image. The technique is based on scale shape analysis, which in turn is based on the assumption that, characters have line-type shapes locally and blob-type shapes globally. In the scale shape analysis, Gaussian filters at various scales blur the given image and larger size shapes appear at larger scales. To detect these scales the idea of principal curvature plane is introduced. By means of normalized principal curvatures, characteristic points are extracted from the scale space x-y-t. The position (x, y) indicates the position of the figure and the scale t indicates the inherent characteristic size of corresponding figures. All these characteristic points enable the extraction of the figure from the given image that has line-type shapes locally and blob-type shapes globally. Dong et. al. [10] has presented an Histogram based approach for extracting a plate. Kim G.M et. al.[9] used Hough transform for extracting the licence plate. The algorithm in this method consists of 5 steps. In the first step the gray scale image is threshold to give binary image. In the second stage the resulting image is passed through two parallel sequences, in order to extract horizontal and vertical line segments respectively. The result is an image with the edges being highlighted. Coming to the third step the resultant image is then used as input to the Hough transform. this produces a list of lines in the form of accumulator cells. In fourth step, the above cells are then analyzed and line segments are computed. Finally the list of horizontal and vertical line segments is combined and any rectangular regions matching the dimensions of a license plate are kept as candidate regions. The disadvantage is that, this method requires huge memory and is computationally expensive. Park et. al. [11] devised a method to extract Korean license plate depending on the colour of the plate. A Korean license plate is composed of two different colours, one for characters and other for background and depending on this they are divided into three categories. In this method a neural network is used for extracting colour of a pixel by HLS (Hue, Lightness and Saturation) values of eight neighbouring pixels and a node of maximum value is chosen as a representative colour After every pixel of input image is converted into one of the four groups, horizontal and vertical histogram of white, red and green (i.e. Korean plates contains white, red and green colours) are calculated to extract a plate region. To select a probable plate region horizontal to vertical ratio of plate is used.

2.5 Segmentation:

This section discusses about the previous work done for the segmentation of characters. Image segmentation may be defined as the partitioning of an image into objects or regions that have coherent characteristics and are meaningful for application at hand. Many different approaches have been proposed in the literature and some of them are as follows, Nieuwoudt et. al. [8] used region growing for segmentation of characters. The basic idea behind region growing is to identify one or more criteria that are characteristic for the desired region. After establishing the criteria, the image is searched for any pixels that fulfil the requirements. Whenever such a pixel is encountered, its neighbours are checked, and if any of the neighbours also match the criteria, both the pixels are considered as belonging to the same region. C. Oz and F. Ercal et. al.[8(a)] have used projections and binary algorithm for character segmentation. Breemananth R &chitra[3] A proposed a system of bob colouring



algorithm and constraint checking for character segmentation.. Morel et. al. [7] used partial differential equations (PDE) based technique, Neural network and fuzzy logic were adopted in for segmentation into individual characters.

2.6 Recognition:

This section presents the methods that were used to classify and then recognize the individual characters. The classification is based on the extracted features. These features are then classified using either the statistical, syntactic or neural approaches. Some of the previous work in the classification and recognition of characters is as follows, Hasen et. al. [23] discusses a statistical pattern recognition approach for recognition but their technique found to be inefficient. This approach is based on the probabilistic model and uses statistical pattern recognition approach. Cowell et. al. [24] discussed the recognition of individual Arabic and Latin characters. Their approach identifies the characters based on 15 the number of black pixel rows and columns of the character and comparison of those values to a set of templates or signatures in the database. Cowell et. al. [22] discusses the thinning of Arabic characters to extract essential structural information of each character which may be later used for the classification stage. Mei Yu et. al. [18] and Naito et. al. [12] used template matching. Hamami et. al. [25] adopted a structural or syntactic approach to recognize characters in a text document, this technique can yield a better result when applied on the recognition of individual characters. This approach is based on the detection of holes and concavities in the four directions (up, down, left and right), which permits the classification of characters into different classes. In addition, secondary characteristics are used in order to differentiate between the characters of each class. The approaches discussed in this paragraph are based on the structural information of the characters and uses syntactic pattern recognition approach. Hu [1] proposed seven moment that can be used as features to classify the characters. These moments are invariant to scaling, rotation and translation. The obtained moments acts as the features, which are passed to the neural network for the classification or recognition of characters. Zernike moments have also been used by several authors [2] for recognition of characters. Using zernike moments both the rotation variant and rotation invariant features can be extracted. These features then uses neural network for the recognition phase. Neural network accepts any set of distinguishable features of a pattern as input. It then trains the network using the input data and the training algorithms to recognize the input pattern (In this case characters).



2.7Summary:

This chapter reviewed topics relevant to the licence plate recognition system. The relevant and previous techniques used in the main four phases of the system were discussed in here. some of the developed systems are also presented in this. In the first case that is image acquisition a sensing system using two charge coupled devices along with a prism gives better input to the system. As the sensing system covers the wide illumination conditions from twilight to noon under sunshine, this system is capable of capturing images of fast moving vehicles without any blur. In the next phase of licence plate extraction, Hough transform was used to extract the licence plate by using and storing the horizontal and vertical edges information and it has also got a disadvantage as it needs huge memory and it is more expensive computationally. various segmentation techniques and recognition techniques were presented and discussed which have been developed commercially.



Chapter 3.Software Development & Image processing in Matlab:

3.1 Introduction:

This is the introduction on handling images in Matlab Software. When it comes to image processing using Matlab there are many things to keep in mind like using the right format, loading an image, reading an image, how to display an image, keeping the saved data in different types, writing an image, converting different type of image formats. Here we discuss some of the functions (commands) used for this type of operations. Matlab has got a ‘Image processing toolbox’ which is necessary when working with images to use these commands. This tool box is installed in the full version of Matlab software and to check whether it is installed and available, type “ver” in the Matlab prompt. When that is typed we can see all the tool boxes that are installed and available to use on your system. You can use Matlab’s help browser option if you have any further problems and references. Matlab has got many easy commands to process many types of functions very fast and easily. You can have access to online manual for image processing toolbox and any kind of tool boxes which you can access through Matlab help browser.

3.2 Digital image:

This section discusses about the properties of the digital images, the file formats, representation of an image in Matlab version and its functions.

An image is a 2D array of values representing light intensity. in case of image processing the term image represents or refers to a digital image. Image is a function of light intensity. Ex: f(x,y) where f is the brightness of the point(x,y), and x&y represents the spatial coordinates of a picture element. By convention, the spatial reference of the pixel with the coordinates (0,0) is located at the top, left corner of the image as shown in the figure below.

Fig 3.1: Reference of the (0,0) pixel.

3.3 Images in Matlab:

There are many types of image formats, most of the formats are accepted by Matlab.The image formats that are supported by Matlab are:

JPEG PCX TIFF BMP HDF



XWB

The mostly used format for an image is JPEG- images and this is the main format whose images are found on internet. It is the most widely used compression standards for images. When you store an image you can see from the suffix in what format it is stored. For example an image is stored as “plate.bmp” is stored in the bmp format and we can see later on how to load an image of this format into Matlab. If an image is stored as “numberplate1.jpg” then we will see that it is stored in jpg format so that we will be able to load an image into Matlab of this format.

3.4 Working formats in Matlab:

If an image is stored as a JPEG-image format on your disc or drive we must first read into Matlab. Whenever we started working with an image or to perform any operation on image like performing a wavelet transform on the image, we need to convert the image into a different format. Here we discuss some of the formats.

3.4.1 Binary imageThis image format also stores an image as a matrix but can only colour a pixel black or white (and nothing in between). It assigns a 0 for black and a 1 for white.

3.4.2 Intensity image (Gray scale image)Intensity image is almost equivalent to “gray scale image” and we mostly use this type of image. It is a matrix in which every element will have a value relative to how intensively pixel is in the corresponding position must be coloured. In two ways the number can be represented which will gives the details of the pixel intensity. the first one is data type or it is called as double class which is used to assigning a floating number between zero to one to every pixel. Zero means black colour and white implies to 1. Second class is unit8 which will assign a number or integer from 0 – 255 to represent the intensity of the pixel. The same way 0 implies to black and 1 to white. This class requires nearly 1 by 8th of the storage when compared with the first class. But most of the mathematical functions can be applied in the first type of class that is double class.In this thesis the work is mostly done with gray scale images. when we can work with gray scale and learn that then we can know the principle of working with colour images.

3.4.3 Indexed Image:This represents colour images in a practical way. Indexed image stores image as two matrices. In the first matrix it has the same size as image and one number for each pixel and in the second matrix its size may be different from image and it is called as colour map. The numbers in the first matrix is an instruction of what number to use in the colour map matrix.

3.4.4 RGB imageThis format is for colour images. In this type of format an image is shown with 3 matrix with the same sizes that match format of the image. In this every matrix implies to 1 of colours in red, green or blue. The matrix give instructions to pixels about the certain amount of these colours that should be used.

3.4.5 Multi frame image



Multi frame format is good for sequence of images. there are some areas where we may need to work with the image sequences. This type of image is often used in medical and biological imaging..

3.5 Conversion between different formats:

When working with images in Matlab there will be a need to change the image formats. Below is examples of some conversion functions that could be found in Matlab.(Within the parenthesis you type the name of the image you wish to convert.) m2bw() - converts image to binary image based on threshold

• graythresh -Global image thresholding • Ind2gray() - converts indexed image to gray scale image• dither() - converts intensity/indexed/RGB format to binary format• gray2ind() - converts intensity format to indexed format• ind2rgb() -converts indexed format to RGB format• mat2gray() -converts a regular matrix to intensity format by scaling• rgb2gray() -converts RGB format to intensity format• rgb2ind() - converts RGB format to indexed format

All the commands mentioned above needs image processing tool box for them to perform their operations.

3.6 Reading & writing image files :

When an image is considered which you want to work with, it is usually in a file format like when you download an image from the web it is usually stored as a JPEG file. When the image processing has completed, we may want to write it back to a JPEG-file so that we can, like posting the image on the internet.To do these things there are commands meant to perform these operations. they are imread and imwrite. These commands also require image processing toolbox.

→ imread():This command is used to read an image. The name of the file to be read has to be kept within the parenthesis. The name of the file has to be in single quotes ‘ ‘.

→ imwrite( , ):This command is used to write an image to a file. The name of the image you have worked with has to be typed within the parenthesis as the first argument. as a second argument within the parenthesis type the name of the file and format that you want to write the image to. The name of the file has to be kept within the single quotes ‘ ‘. Semi-colon should be used after these commands, otherwise lots of number scrolls on the screen.

3.7 Saving variables in Matlab:

This is about saving and loading variables in Matlab. When a file is read it has to be converted into an intensity image. After converting and worked with the matrix we may need



to save the matrix which represents the image for next time. This can be done using the load and save commands.Commands & operations:Save x: it saves the variable xLoad x:it loads the variable x

3.8 Displaying image in Matlab :

Here we discuss about couple of basic Matlab commands used for displaying an image. These commands do not require any tool box.

→ imagesc(x):Display an image represented as the matrix x.

→ brighten (s):it adjusts the brightness. S is a parameter such that -1<s<0 gives a darker image, 0<s<1 gives a brighter image.

→ colormap(gray):it changes the colours to gray.

There will be times when image is not displayed in gray scale though after conversion to grayscale. Then colormap(gray)command can be used to force Matlab to make that image to be displayed in gray scale. When Matlab is used with image tool processing box then it is recommended to use the imshow command to display the image.

Displaying image on matrix form:There are some commands to display an image that is given on matrix form. It requires image processing tool box. → imshow(x):Display an image represented as the matrix x.→ zoom on:it is used to zoom in using the left and right mouse button→ zoom off:it is used to turn off the zoom function.



4.Chapter:Requirement Analysis:

4.1 Introduction:

The automatic Licence plate localization and recognition system has become an important application of artificial intelligence. This technology is gaining popularity in traffic and security installations. This technology can use existing closed-circuit television or road-rule enforcement cameras, or ones specifically designed for the task. This technology may tend to be region specific, owing to licence plate orientation from place to place. An efficient licence plate recognition process may become the core of fully computerised road traffic monitoring systems, surveillance devices, safety supervision systems and electronic fee collection solutions. It is also important that the recognition accuracy of such a process is very high.

This chapter firstly gives introduction and presents the applications and uses of a licence plate recognition system. In the next stage we discuss the requirements and the elements required for the system, following this typical working of recognition system is discussed. Next the structure of proposed system is presented. At last this chapter ends with the brief overview of the rest of the thesis.

4.2 Applications of Recognition system :

Licence plate localization and recognition system is one form of automatic vehicle identification system. Because of their applications they are of considerable interest. This system has potential applications in areas like vehicle authentication, highway toll collection, parking lot management, border surveillance, security, speed limit enforcement etc.It plays a major role in traffic monitoring systems and maintaining law enforcement on public roads. This area is very challenging because it would need many computer vision solutions like object detection and character recognition. There are many applications of recognition systems. Some of them are briefly described below.Law enforcement:When a vehicle breaks the traffic law the licence number plate can be used to produce a violation fine on the vehicle if the vehicle illegally uses the bus lane or if the vehicle is speeding than allowed. This can be used to detect the stolen or lost vehicles. This system has gained popularity in traffic applications as there is no need to install any additional tracking apparatus. The main advantage of this system is that the system can store the image for future records and reference. the rear part of the vehicle is extracted off the filmed image and is processed after that. The result will be given as input to the database record. The owners of vehicles who violated the law will be able to pay the bill through online with the image of car as proof.Some of the examples where this system is used in traffic monitoring systems are

UK has an extensive number plate recognition CCTV network. Effectively, the police and security services track all car movements all over the country and able to track any car at any time.

Vehicle movements are stored for 5 years in National data centre analyzed for intelligence and to be used as evidence.

Ontario’s 407 ETR highway uses a combination of recognition system and radio transponders to toll vehicles entering and exiting the road



Border surveillance:This is very important application in many countries to protect the border from allowing unauthorised vehicles into the country. It can be used to monitor the border crossings. Each vehicle information will be stored in the database and if any other unauthorised vehicle comes then it assists the registry of entry and exits to a country. Vehicle authentication:This system is used as application in allowing the authorised vehicle into the secured area and stopping the unauthorised vehicles from entering into that area. All the data of the desired area vehicles are stored and when an unauthorised vehicle comes it will stop the vehicle. Parking:The recognition system is used to collect the park fee automatically. It enters the pre-paid members automatically to calculate the parking fee for non-members by comparing the entry and exit timings. The licence number plate is recognised and stored and upon its exit the car plate is read again and the driver is charged for the duration of parking.

Example: In Stockholm, Sweden, the system is used for the congestion tax, owners of cars driving into or out of the inner city must pay a charge, depending on the time of the day.

Toll fee gates:Normally if the toll gates are operated manually then the vehicle has to be stopped manually and collect the required tariff but in order to decrease the manual power and to be efficient the automatic toll gates are developed. In the automatic toll gate system the vehicle would no longer need to stop because as the vehicle passes the toll gate it would be automatically classified in order to calculate the correct tariff.

Examples: Many cities and districts have developed traffic control systems to help monitor the movement and flow of vehicles around the road network like

Homeland Security:- These licence recognition systems can be used to keep a tight watch on entire cities, ports, borders and vulnerable areas if they are fixed at a certain place. Fixed system scan be mounted to bridges, gates and other high traffic areas. This system allows a command centre to organise and strategic efforts in reaction to the information captured because of its ability to read strings of alpha-numeric characters and compare them instantaneously to Hot lists.Every acquisition system like camera will capture critical data such as colour photos, date and time snaps, as well as GPS coordinates on every vehicle that passes. This database provides very important information and clues of proof which can greatly aid law enforcement with

Pattern recognition Placing a suspect at a scene Watch list development Identifying witnesses Visual clues revealed within the image of a car’s environment

This recognition system can also be used for the following applications Car park usage Pedestrian crossing usage Number of vehicles along a road



Areas of low and high congestion Frequency , location and cause of road works

4.3 Required elements for system:

The typical recognition system consists of the following units:The basic plate recognition system consists of a camera, frame grabber, computer and specially designed software for image processing and analysis.

Digital Camera: is used to take image of a vehicle from front side or from rear end.

Frame grabber: acts as an interface board between the camera and the PC that allows the software to read the image information.

Computer:A normal PC running with windows or Linux is used to run the application that controls the system, reads the image, analyzes and identifies the plate and interfaces with other applications and systems.

Illumination:A controlled light that can bright up the plate,and allows day and night operation. In most cases the illumination is Infra-Red(IR) which is invisible to the driver.

Software:It is the important element required for a system to process the application and recognition package.

Hardware:This element is used to act as a interface between the various input/output boards and the external world like control boards and networking boards.Database:The events are recorded in local database or transmitted over the network. The data in that includes recognition results and the vehicle or drive image file.The efficient system should be capable of-working indoor and outdoor- working in a wide range of illumination conditions- being invariant to size, scale and stroke thickness- being robust to broken strokes, printing defects, noise, etc.- being robust to camera-car relative movement- giving a real-time response



4.4 colour spaces (models):

Commercial LPLR systems usually rely on monochrome acquisition systems and the character segmentation is achieved via binarization and connected component analysis. Optimal thresholds are computed based upon local or global histogram analysis. The pixels above the threshold are considered as foreground and the rest as background. fig 3.2 shows special licence plate examples. In order to simulate the binarization of monochrome images, the input colour images are converted to gray scale by preserving the intensity component as described by the HSI colour model which is discussed in the next chapter.

Fig 4.1 several UK licence plates . Several special UK license plates are displayed in colour in the top row. The second row shows the result of binarizing the corresponding gray-scale (intensity) images, based upon a global threshold.

The choice of a colour or feature space is of great importance, since the ability to discriminate text from background depends upon the coordinate system into which the license plate image is projected. A colour space, also called a colour model, is a specification of a coordinate system. Many specialized models are available due to the wide range of applications of colour science. Two different colour models were considered for experiments in this section: RGB&HSV/HSI.

4.4.1 RGB colour space:The colours that we mostly perceive may be defined as additive mixtures of three fixed primary stimuli in appropriate proportions according to the trichromatic generalization (Wyszecki and stiles, 1982). So it is convenient enough to represent the colour stimuli by vectors in a 3D space RGB cube is a common hardware-based model in which the device tristimulas colours are driven directly by vector components.The RGB model forms a cube with x, y and z axes as G,B and R respectively based upon a left-handed Cartesian coordinate system as shown in the figure 3.3.



Black is at the origin while white is at the corner farthest from the origin. The line connecting these two points represents the gray scale where r = b = g. Primary colours (red, green & blue) and secondary colours (cyan, magenta and yellow) are other significant corners. RGB values are usually normalized to be in the range [0, 1], instead of the range [0, 255] commonly used to represent a colour value from digital camera. To represent all 3 colour values, each pixel needs a total of 24 bits which RGB model allows for 16,777,216 different colours. Fig. 3.9.1(b) shows an approximate representation of these colours in three exterior surfaces of the RGB cube.

.(a) (b)

Fig. 4.2 (a) The RGB coordinate system schematic, after Gonzalez and Woods (2002). (b) A colour view of the RGB cube

4.4.2 HSV and HSI colour spaces:RBG model is good for supporting hardware but it is not natural for humans. Let us consider an example. it is difficult to understand how to “lighten” a colour in RGB space.Coming to hue, saturation and brightness they are more comfortable to humans than RGB as hue represents the experience of pure colour in terms of dominant wavelength like red, orange and yellow etc. Saturation describes the degree to which a pure colour is undiluted by white light like pink is diluted red. In other words saturation is the dominance of hue in the colour Brightness describes overall intensity or strength of light ( e.g., dark pink vs light pink). The HSV (hue, saturation, value) and HSI (hue, saturation, intensity) colour models make use of these human response descriptions to represent colour information. An advantage of both models is that the brightness or intensity component is separated from the colour information, which is represented by the hue and saturation components. We are interested in a colour space that better describes human colour perception because license plates are designed with



human readability in mind. Therefore, the colour characters and background will be easier in a colour or measurement space that expresses dissimilarities in colour (or distances) as perceived by human beings.

Fig: 4.3 HSV Schematic.

4.5 Optical character recognition:

The three basic principles integrity, purposefulness and adaptability constitute the core of OCR allowing it to replicate natural or human-like recognition. It provides machine vision functions we can use in an application to perform OCR.

4.5.1 Introduction:OCR is a type of document image analysis where a scanned digital image that contains either machine printed or handwritten script is input into an OCR software engine and translating it into an editable machine readable digital text format. It is the process by which the machine vision software reads text and/or character in an image.

4.5.2Applications:Machine vision OCR is typically used in automated inspection applications to identify or classify components. For example, OCR can be used to detect and analyse the serial number on an automotive engine that is moving along a production line.We can use OCR in wide variety of other machine vision applications like, they are used for inspecting pill bottle labels and lot codes in pharmaceutical applications. They are used foe sorting and tracking mail packages and parcels. They are used in verifying wafers and IC package codes in semiconductor applications. They are also used for reading alphanumeric characters on automotive parts.



Fig: 4.4 Licence plate recognition using OCR software

4.5.3 Training characters:Training involves teaching OCR the characters and patterns that are to be detected during the reading procedure. Figure 4.5.3 illustrates the steps involved in this process. The process in which the characters are located in an image is referred to as character segmentation. When we finish segmenting the characters, we’ll use OCR to train the characters, storing information that enables OCR to recognize the same characters in other images. We train the OCR software by providing a character value for each of the segmented characters, creating a unique representation of each segmented character. You then save the character set to a character set file to use later in an OCR reading procedure.

Fig: 4.6 Steps of OCR reading procedure



4.6 Thresholding:

Thresholding is one of the most important concepts in the segmentation process. It is the process of separating image pixels into foreground and background pixels based on their intensity values. The pixels whose intensity values lie outside the lower and upper threshold values of the threshold range are called background pixels. Foreground pixels are those whose intensity values are within the lower and upper threshold values of the threshold range.OCR includes one manual and three automatic methods of calculating the threshold range .Fixed range is a method by which we can manually set the threshold value uniform, linear and non-linear are three automatic methods of calculating the threshold range.

4.7 Algorithms:

There are six primary algorithms that the software requires for identifying a licence plate: Plate localisation – responsible for finding and isolating the plate on the picture. Plate orientation and sizing – compensates for the skew of the plate and adjusts the

dimensions to the required size. Normalisation – adjusts the brightness and contrast of the image. Character segmentation – finds the individual characters on the plates. Optical character recognition. Syntactical/Geometrical analysis – check characters and positions against country-

specific rules.During the 3rd phase some systems may use edge detection techniques to increase the Picture difference between the letters and the plate backing. A median filter may also be used to reduce the visual noise on the image.



Fig:4.7 Flowchart of licence plate localization algorithm.

To get the perfect and good quality of the licence plate localization software the algorithm should be good. If better algorithm is developed the results also will be good. For a better algorithm it will have

Highest quality of recognition software Can handle many type of plates Fast processing speed Widest range of picture quality can be handled Most tolerant against distortions of input data

A good algorithm should also be capable of reading all plates with the same quantity level. For example let us consider the plates in the Europe, so there are various types of licence plates in Europe like

Black or dark characters on white colour plate White characters on black colour plates 1- row type plates 2- row type plates Different sizes of characters on plate Latin fonts Plates without any region mark or shield

The acquisition phase of this technology determines the image with average quality with which the licence plate localization algorithm has to work. As we know that if the quality of



the input image is good then algorithm will definitely have good conditions and hence the accuracy of the licence plate recognition will also be high.In order to expect good results from a licence plate recognition algorithm, the images that are processed must contain a good

Spatial resolution Sharpness Contrast Lighting conditions Position Angle view

All the images should posses above qualities so that good results will be obtained. Here some plates are shown which are problematic

(a)

(b)

(C)

(d)

(e)

Fig:4.8 problematic plate images

(a)This image has too small characters on the plate which leads to low spatial resolution.

(b)This image is blurred so the characters cannot be recognised properly.

(c)This image has very low contrast.

(d)This image has over lighting or bad lighting conditions.

(e)This image is highly distorted.



An image acquisition system is considered to be good if it provides a stable, balanced, reasonable good image quality under all of its working conditions.

4.8 Working of system:

A Licence plate localization & recognition system can be considered conceptually as having two separate processing steps:

Licence plate localization Licence plate character recognition

To describe the working of the system let us consider an example figure and discuss the working of it.

Fig: 4.9 an illustration of working system

As shown in the above figure, when the vehicle comes to the given secured area the recognition system unit (as shown) will recognise the car and then it will activate the illumination that is invisible infra red which is used in many cases. The system takes the picture using camera from two sides that is from front end and rear end in order to cover the whole vehicle. The image taken will have the number plate of the vehicle as the picture is taken from both the ends. Here the main process starts where the image having number plate is fed to the system as an input image after which the system enhances the image, detects the position of the plate, extracts the licence plate, segments the characters on the plate and then recognises the characters on the plate.After all the process is done the system checks with the stored database whether the vehicle is an authorised or unauthorised one. If the vehicle is found on the predefined list of authorised vehicles the system will open the gate and allows the vehicle to go on.

The system can be setup in a way to make a welcome note or show some green or red light etc. When the vehicle entered into the secured area the gate will be closed and the process continues. This is how the working of a recognition system is carried on.



4.9 Structure of proposed system:

The proposed is developed in two models (A)Colour extraction method(B) Edge detection and extraction method.

4.9.1 colour extraction method(1 st method): The implementation of the program is developed in Matlab.The steps in this model of extraction of licence plate is

• Pre processing• Coloured plate extraction• Extracting the number plate

Assumptions: Input is an image of a stationary Car. Only the most common type of license plates i.e., single line plates will be dealt with. The license Plate has a yellow background with text written in Black.

Pre processing : We have assumed that the licence plate has a yellow back ground, firstly we have to identify the regions of the image in which it is containing the intensity of RGB corresponding to the colour Yellow coloured part from the images are filtered using the values obtained

- (a< R< b) && (p< G< q) && (x< B< y)Where R is the intensity of the red colour, B is the intensity of blue colour and G is of the green colour. By using this basis we get a binary image. We change the yellow to white and non yellow to black.Extraction:We can use morphological operations also here to get the number plate extracted from the image. But here we are using another process of colour mapping to RGB colours. As these are the primary colours they are enough to compose any type of colour adequately even though the illumination is infinite dimensional. The YIQ representation is the alteration to the RGB representation where Y is the luminance or brightness which imply to colour density, I is the hue which is the dominant such as orange, red or yellow, Q is the depth or saturation, it is the amount of white light mixed with a hue of colour.The RGB colour space is correlated so it is not suitable for image processing applications, but other models like YUV, HIS are suitable for image processing applications.

Removing background for plate extraction:To get a accuracy of extracted plate from the image the background removal is one of the best process. Below a flow chart shows the methodology for getting a gray image from the image of vehicle.



Fig:4.10 Background removal flowchart

Background subtraction is one method which is popular and effective for detecting moving objects in a scene. Based on the probability concept the background image can be constructed from the histogram which is modified of individual pixel in image sequence.Figure below shows the original image and the next image shows the extracted number plate after the background colour is removed with black background.

Fig:4.11(a) original image (b) extracted number plate

There are still some noise distortions which can be seen on the background removed image. This noise can be removed using a median filter and if the background colour is removed and replaced by white colour then the executed will be in the way as shown below. The picture below shows the image of original image taken and the image is fed to the input of the system to extract the licence plate from the original image by using background colour method. Original plate and the figure after undergoing the process of colour method the resultant image will be as shown below

(a)Original plate (b) extracted plate



Fig:4.12(a) original image (b) extracted number plate

The type of method we discussed above is colour reduction method which we have used to extract the licence number plate from a given image but this system is not perfect and efficient as we have so many assumptions like the colour of the plate and the angle in which the image has to be captured.

Problem:This system is not applicable for all type of number plates. There are situations when this system cannot be applicable and the areas where this system is not applicable is discussed and shown in the next chapter of experiments and results. An example of such type is shown below. The below is the image of the vehicle is processed in the same way the previous plates are processed and the resultant image is shown below.

(a) (b) Fig :4.13(a) original image (b) extracted number plate

The above image is not efficient enough to extract the licence plate from the image and to overcome this assumptions and all these problems we have proposed a edge detection and extraction method which is good efficient than the previous approach. The edge detection and plate recognition method id discussed below.

4.9.2: edge detection and extraction method:(2 st method) The proposed system is designed to recognise the licence plate from the front end and rear end of the vehicle. The input to the system is the image of the vehicle acquired by a digital camera and the output of the system is the localization and recognition of the extracted licence plate.

The implementation of the program is developed on MATLAB. The step wise approach is:

Pre processing (Image Acquisition) Plate Extraction Character Segmentation



In the first phase of the system the image is acquired. In the second phase the system extracts the licence plate region from the image. The third phase isolates the characters, numbers and letters.

Fig 4.14: licence plate recognition systems:

4.9.3 pre-processing :The input image consists of many colours and the image is processed initially to improve the quality and prepares it to next phases of the system. Since the image has different colours the system will convert the RGB images to gray scale images using NTSC standard method. Gray=0.299*Red+0.587*Green+0.114*BlueIn the next phase the gray image is filtered using median filter in order to remove the noise, while preserving the sharpness of the image. The filter used is a non-linear filter where it replaces each pixel with a value obtained by computing the median of values of pixels.

4.9.4 Plate extraction:The plate extraction is the second phase and the most important phase of the recognition system. The plate extraction process consists of 5 parts as shown below. In this each phase performs a process of segmentation on the gray scaled image to eliminate the pixels which does not belong to the licence plate region.Let us consider an example that the horizontal localization phase is responsible for detecting the horizontal segments that contain the number plate. In the same way vertical localization phase is responsible for locating the vertical segments of the number plate.Let us consider that we are taking into consideration of England number plates. Three types of plates are available

White characters on black background



Black characters on yellow background Black characters on white background

This is just a consideration to explain the meaning of the edges in the image, for example if we take a licence plate consist of a row of white characters on black background, we can say that the licence plate is characterised by a row of transitions from black to white and vice versa, these transitions are called edges.

Fig: 4.15 number plate extraction phase

In this thesis we use Sobel operators to find the edged image. The sobel command performs a 2 dimensional spatial gradient measurement on an image. Normally sobel operator is used to find the approximate absolute gradient magnitude at each point in an input image which is the gray scale image. The actual sobel masks are shown below.

(a) (b)

Fig 4.16: sobel masks for edge detection (a) vertical (b) horizontal

After creating the image with edges the system will try to find out the regions with high edge values because the high edge values are most likely contain a licence plate. In order to that a horizontal projection profile is constructed by the system. The vector sum of the pixel intensities in each row is defined as the horizontal projection profile.



(a)(b)

Fig 4.17: (a) normal image (b) sobel edge detection image

After finding the horizontal position of the plate the next step is to find the vertical position of the plate. For vertical projection profile, the methods usually use a statistical study of histograms as discussed.

Fig: 4.18 Horizontal projection of edge image

To locate the vertical coordinates of the plate using image edge a simple model is proposed. The first step in the algorithm is to slide a window through each horizontal segment and adding the values of edges inside window and then dividing the result by windows area to get magnitude. in the same process for each horizontal segment the algorithm stores the value of each window in a new vector for next processing i.e. HDV.



Fig: 4.19 vertical projection analysis.

After both localization processes the next step is to isolate the licence plate from any redundant background after which the skew angle detection and corrections are made. The isolation of the licence number plate from any type background can be done using a projection profile that would reflect the number of white pixels in each row and column. An example is shown below

Fig: 4.20 horizontal and vertical projection of licence plate.

Fig: 4.21 vertical localization of licence plate.



4.10 Segmentation:

In our proposed model character segmentation is not done but we shall discuss briefly the process that can be implemented after the above process has been done. It is the process of extracting the licence plate and the numbers from the image taken. There are different aspects that make that make this concept little complicated like noise in the image, frame of the plate, plate orientation, light intensity and space marks. Many systems have been proposed to overcome these problems. The method that is suggested in this thesis after result of the proposed system is1)Pre processing which includes

Converting image to gray scale binarization

2) object enhancement algorithmThe object enhancement algorithm consists of two steps.

Firstly, gray level of all pixels is scaled into the range of 0 to 100 and compared with the original range 0 to 255, the character pixels and the background pixels are both weakened.

Secondly, sorting all pixels by gray level in descending order and multiply the gray level of the top 20% pixels by 2.55. Then most characters pixels are enhanced while background pixels keep weakened.



Fig: 4.22 object enhance method (for character segmentation)

The above figure shows the result of object enhancement.3) After pre processing and object enhanced algorithm Horizontal segmentation is done and vertical bounds are noted to segment the characters on the number plate of the image.



Chapter5: Proposed solution & experimental results

5.1 Problem definition:

In a developing are developed country traffic is a significant problem. Huge integration of Information technology into every aspect of the modern life shows the demand for processing of vehicles as conceptual resources in information systems. A special and intelligent equipment was needed which is able to recognise the vehicles by their number plates in real environment and reflect it to conceptual resources. As a result of this various systems are developed and today there are many places where those systems play a vital role like traffic, security applications, border control and parking etc. Till now many systems are developed in neural networks, using vision assistant and lab VIEW and Matlab. This work is also done using Matlab.

5.2 Proposed solution:

For real time applications, the system would require a video camera which captures the image of vehicles from both the rear and front ends. For the present work also sony camera (frame_grabber) is used to acquire the image.

The images of the different vehicles are captured manually and after that they are fed to the software where they are first converted in to gray scale images. After this step horizontal segmentation is done through the image to find the regions of number plate and thereafter vertical segmentation is done after which the licence plate is localised and extracted.

5.3 Results:

To test the proposed system experiments have been performed. The simulation process is carried out in Matlab 7.1 for the licence plate localization and recognition. A set of 50 images were used for testing the proposed system. The images were taken from various environments. They are of different sizes and different colours. The plates were taken from complex images which contain different colours and several objects in the image. The images are taken from different inclined angles and distances relative to camera.

Some of the examples which are successful for coloured method are given below. the normal images are fed to the software to get the gray image where numbers or letters can be identified after which the licence plate number is located and extracted.

The below table shows the original images of the vehicles which are used for experimenting the results.



Fig 5.1 original image plates used for experiment



Example 1:

(a) (b)

Fig 5.2 (a) original image (b) extracted number plate

Example 2:

(a)

(b)

Fig 5.3(a) original image (b) extracted number plate



Example 3:

(a)

(b)


Example 4:

(a) (b)




Example 5:

(a) (b)


Example 6:

(a) (b)


PROBLEMS: There are some problems where this system will not work. Some of such examples are shown. The system fails in the first step of extracting the gray scale image where the numbers can be located. Below are some example plates where the system fails



Fig:5.8 Four different examples where the first proposed system fails

The software fails to give the correct results for the above images



The above shown images are examples.

In the above shown example most of the images are having background which is problematic. If you see the 2nd and 4th images they are having more distortions in the image and half of the images are not covered by the vehicle. Since the image of the vehicle is not the important part in the image, that will be a problem to extract the licence plate without any distortions. If the early images are considered then there are some images where there are more numbers and words other than number plate. If the plate is in such a way then the extracted plate would contain those number and words which are not in the extracted plate. In order to overcome these problems edge detection and extraction method is proposed.

Results for proposed second method:

Since the first method that we have approaches is not efficient one, we have proposed the edge detection and extraction method. There are sample examples shown below for this method. Some of the original images that are used for the experiments are shown below. These are some sample examples.

Fig:5.9 original images used for sample experiment



Example 1:

(a) (b)

(c) (d)

(e)

(f)

Fig: 5.10.(a) Gray scale image (b) edge density image (c) projection graph (d) licence plate location (e)detected plate area (f)extracted number plate



Example 2:

(a)

(b)

(c)

(d)

(e) (f)

Fig: 5.11 (a) Gray scale image (b) edge density image (c) projection graph (d) licence plate location (e)detected plate area (f)extracted number plate



Example 3:

(a) (b)

(c) (d)

(e)

(f)

Fig: 5.12(a) Gray scale image (b) edge density image (c) projection graph (d) licence plate location (e)detected plate area (f)extracted number plate



PROBLEMS: In the same way the above process is carried out it has been done on many vehicle plates and there is 92% accuracy in the success of the system but still there are some number plates where the system fails to extract the correct licence plate and which is difficult for further process like segmentation and recognition. There are some examples shown below where the system fails to get the correct extraction of the licence number plate. The reason why the system fails in this situation is explained below.



Sample example where the system fails :

(a)

(b)

(c) (d)

(e)

(f)

Fig: 5.13(a) original image, (b) gray scale image, (c)edge density image, (d) projection graph (e)locating the different place instead of number plate ,(f)extracted area is not a number plate.



Since the image is not quite clear and because of many edges in the image the exact number plate cannot be extracted. The extraction of the image can be problematic due to the weather conditions also as if the image is not clear enough to see then the plate cannot be extracted. Here in this image the clarity of the image is also not good.

5.4 Problems encountered:

There are some main problems that are faced in my project and some general problems that occur during the process of licence plate localization and recognition are

The software will face difficulty to localize and recognise the licence plate if the size and the format of the plate changes. If it is only slight difference then the software can overcome but if the total system, like the number plate, alphabets and the colours are different then it will be a problem for the software.

If the license plate happens to be too much tilted from horizontal, then the result of segmentation of the license plate will be poor.

In the image processing system, the systems are greatly affected by the lighting conditions. Here our proposed system is also sensitive to lighting conditions. Lightening conditions should be kept constant when working in the middle of the road.



Chapter 6: Conclusion and future scope

6.1 conclusion:

What is normal and easy for human eye may seem difficult task for the computer. but computer has a powerful vision which has a powerful tool that provides the capabilities of performing very useful operations as the one we used in this project. The process of number plate recognition system needs very high accuracy when working on busy road or parking area which may not be possible manually when the traffic is heavy. We cannot keep track of vehicles when there are multiple vehicles are passing in a very short time and we the humans het fatigued due to monotonous nature of the job. In order to overcome this problem, many experiments were done and many efforts were made to develop an accurate number plate recognition system. Here in our project our purpose is to design a system for number plate localization and recognition where if an input image is given, it should firstly localize the licence plate and then extract the exact licence plate without any distortions. The proposed system will search for image with high density edge regions which may contain number pate, after that process filtered process is carried out to extract the regions that contain the number plates and filter out the regions that do not contain number plate and after which the number plate is extracted correctly. We got an overall efficiency 0f 90% for the system as the system has been tested with nearly 50 vehicle number plates. Even this efficiency is not acceptable generally, but the system can be used for vehicle authentication and identification. It can be concluded that the proposed system has been by and far successful. For vehicle licence plate localization and recognition using a intel core duo(t6400) 2.2GHz CPU and 2Gb RAM in Matlab 7.1 will tale only 0.3 second to localize a licence number plate from busy traffic by this proposed algorithm.

6.2 Future scope:

There are some aspects that have appeal for the future research. First recommendation for the future work is to acquire large and various database of the number plates. There is a scope in the future where the system can be able to work where the number plate, the colour and the font of the plate is identical with varied font sizes. The system should not compromise and it should be sensitive and should be able to locate the plate at any conditions as tracking stolen vehicles and monitoring vehicles for homeland security cannot be compromised where 100 % accuracy is neededThere are also some situations where the systems fail to get the accuracy in the issues like stains, smudges, variety font styles, blurred regions and the sizes. The future work can be extended to minimize all these errors.



References:

[1] Hu, M. K., "Visual Pattern Recognition by Moment Invariant", IRE Transaction on Information Theory, vol IT- 8, pp. 179-187, 1962.

[2] Khotanzad, A., and Hong, Y.H., "Invariant image recognition by zeraike moments," IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 12, no. 5, pp. 489-497, 1990.

[3] Bremananth, R.; Chitra, A.; Seetharaman, V.; Nathan, V.S.L.,“A robust video based license plate recognition system” IEEEProceedings of ICISIP, 2005, Page(s): 175 – 180.

[4] B. Hongliang and L. Changping, “A hybrid license plateextraction method based on edge statistics and morphology,” inProc. ICPR, 2004, pp. 831–834.

[5] D. Zheng, Y. Zhao, and J. Wang, “An efficient method oflicense plate location,” Pattern Recognit. Lett., vol. 26, no. 15,pp. 2431–2438, Nov. 2005.

[6] Comelli, P., Ferragina, P., Granieri. M. N., and Stabile, F., "Optical recognition of motor vehicle license plates", IEEE Transactions on Vehicular Technology, vol. 44, no. 4, pp: 790-799,1995

[7] Morel, J., and Solemini, S., "Variational Methods in Image Segmentation", Birkhauser, Boston, 1995.

[8] Nieuwoudt, C, and van Heerden, R., "Automatic number plate segmentation and recognition", Seventh annual South African workshop on Pattern Recognition, pp. 88-93, IAPR, 1996[8](a) C. Oz and F. Ercal, “A practical license plate recognitionsystem for real-time environments,” in Proc. IWANN, vol. 3512.Berlin,Germany: Springer-Verlag, 2005, pp. 881–888

[9] Kim, G. M., "The automatic recognition of the plate of vehicle using the correlation coefficient and Hough transform", Journal of Control, Automation and System Engineering, vol. 3, no.5, pp. 511-519, 1997. 75

[10] Cho, D. U., and Cho, Y. Ft., "Implementation of pre-processing independent of environment and recognition and template matching ", The Journal of the Korean Institute of Communication Sciences, vol. 23, no. 1, pp. 94-100, 1998

[11] Park, S. FL, Kim, K. I., Jung, K., and Kim, H. J., "Locating car license plates using neural network", IEE Electronics Letters, vol.35, no. 17, pp. 1475-1477, 1999.

[12] Naito, T. Tsukada, T. Yamada, K. Kozuka, K. and Yamamoto, S., "Robust recognition methods for inclined license plates under various illumination conditions outdoors",



Proceedings IEEE/IEEJ/JSAI International Conference on Intelligent Transport Systems, pp. 697-702,1999.

[13] Naito, T., Tsukada, T., Yamada, K., Kozuka, K., and Yamamoto, S., "License plate recognition method for inclined plates outdoors", Proceedings International Conference on Information Intelligence and Systems, pp. 304-312, 1999.

[14] Naito, T. Tsukada, T. Yamada, K. Kozuka, K. and Yamamoto, S., "Robust recognition methods for inclined license plates under various illumination conditions outdoors", Proceedings IEEE/IEEJ/JSAI International Conference on Intelligent Transport Systems, pp. 697-702,1999

[15] Salagado, L., Menendez, J. M., Rendon, E., and Garcia, N., "Automatic car plate detection and recognition through intelligent vision engineering", Proceedings of IEEE 33r Annual International Carnahan Conference on Security Technology, pp. 71-76, 1999.

[16] Naito, T., Tsukada, T., Yamada, K.s Kozuka, K., and Yamamoto, S., "Robust license-plate recognition method for passing vehicles under outside environment", IEEE Transactions on Vehicular Technology, vol: 49 Issue: 6, pp: 2309-2319, 2000.

[17] Kim, K. K., Kim, K. I., Kim, J.B., and Kim, H. J., "Learning based approach for license plate recognition", Proceedings of IEEE Processing Society Workshop on Neural Networks for Signal Processing, vol. 2, pp: 614-623, 2000.

[18] Yu, M., and Kim, Y. D., "An approach to Korean license plate recognition based on vertical edge matching", IEEE International Conference on Systems, Man, and Cybernetics, vol. 4, pp. 2975-2980, 2000[19] Ioannis pitas “ digital image processing algorithms”.

[20] Yan, Dai., Hongqing, Ma., Jilin, Liu., and Langang, Li, "A high performance license plate recognition system based on the web technique, Proceedings IEEE Intelligent Transport Systems, pp. 325-329, 2001.

[21] Hontani, H., and Koga, T., "Character extraction method without prior knowledge on size and information", Proceedings of the IEEE International Vehicle Electronics Conference (IVEC'01), pp. 67-72, 2001.

[22] Cowell, J., and Hussain, F., "Extracting features from Arabic characters",Proceedings of the IASTED International Conference on COMPUTER GRAPHICS AND IMAGING, Honolulu, Hawaii, USA, pp. 201-206, 2001.

[23] Hansen, H., Kristensen, A. W., Kohler, M. P., Mikkelsen, A. W. , Pedersen J. M., and Trangeled, M., "Automatic recognition of license plates", Institute for Electronic System, Aalhorg University, May 2002.

[24] Cowell, J., and Hussain, F., "A fast recognition system for isolated Arabic characters", Proceedings Sixth International Conference on Information and Visualisation, IEEE Computer Society, London, England, pp. 650-654, 2002.



[25] Hamami, L., and, Berkani, D., "Recognition System for Printed Multi-Font and Multi-Size Arabic Characters", The Arabian Journal for Science and Engineering, vol. 27, no. IB, pp. 57-72, 2002.


faculty: aces msc: full time module: dissertation …hsnemer.weebly.com › uploads › 7 › 6 ›...

Documents