LOCALIZATION OF LICENSE PLATE NUMBER USING DYNAMIC IMAGE PROCESSING TECHNIQUES AND GENETIC ALGORITHMS

Abstract

In this research, a design of a new genetic algorithm(GA) is introduced to detect the locations of the License Plate(LP) symbols. An adaptive threshold method has been applied toovercome the dynamic changes of illumination conditions whenconverting the image into binary. Connected component analysistechnique (CCAT) is used to detect candidate objects inside theunknown image. A scale-invariant Geometric RelationshipMatrix (GRM) has been introduced to model the symbols layoutin any LP which simplifies system adaptability when applied indifferent countries. Moreover, two new crossover operators,based on sorting, have been introduced which greatly improvedthe convergence speed of the system. Most of CCAT problemssuch as touching or broken bodies have been minimized bymodifying the GA to perform partial match until reaching to anacceptable fitness value. The system has been implemented usingMATLAB and various image samples have been experimented toverify the distinction of the proposed system. Encouraging resultswith 98.4% overall accuracy have been reported for two differentdatasets having variability in orientation, scaling, plate location,illumination and complex background. Examples of distortedplate images were successfully detected due to the independencyon the shape, color, or location of the plate.

ARCHITECTURE

EXISTING SYSTEM

Nowadays, speed control stationson highways take color photos of vehicles that break the speedlimit. Usually, on highways in Iran, these control stations uselocal memory because they are not connected to a central database. The photos are taken by speed control camerasin high resolution, which consume a lot of disk space; thus,after a while, most of the stations encounter a low-disk-spaceproblem.

DISADVANTAGE:

1. Low quality

2. Low Speed

PROPOSED SYSTEM

In this system, processing is done on the original image without any resizingor transformation. The license plate is detected by using ageometric template on colonies of target pixels. A colony isdefined as connected component pixels with the same color.The new approach for detection avoids time-consuming algorithms,such as Hough transform, wavelet transform, andcurvelet transform, gray scale or binary conversion in largescale, or even edge detection. Nowadays, due to the developmentsin professional photography equipment and high-speedcamera shutters, it is no longer necessary to carry out high levelpreprocessing for noise reduction and sharpening/deblurringof the original image;

Hence, the preprocessing done at lowlevel ensures faster recognition. Some weather factors (foggyor rainy) and light factors (dark nights or vehicle lights) orbroken, smeared, and unreadable license plates can cause visionproblems, rendering recognition difficult. Using color featuresin the proposed system makes the system more flexible and aidsin generating fairly reliable results in all the weather conditionsbecause of the system’s similarity to human vision anddecision

ADVANTAGE

The license platerecognition (LPR) system can be used in smart parking areasor smart toll stations to open gates for vehicles bearing authorizedlicense plates or to calculate the average speed of avehicle between two stations by recognizing its license plateat both stations. In addition, by installing LPR systems onroads, particularly in traffic zones and at junctions that needpolice patrolling, prohibited vehicles can be recognized and their movement monitored.

MODULES

1. AUTOMATED PLATE RECOGNITION

It is based on developments inimage processing and optimized template-matching methods,proves that a fast LPR system based on color features ispossible and practical. It recognizes license plates in colorimages without any resizing and conversion, thus reducing theresponse time. It can be used in automatic toll stations, tunnels,highways, intelligent parking, and traffic zones for LPR by using surveillance cameras. The results based on images fromspeed control cameras on highways demonstrate the system capability and reliability.

The functioning of the ALPR system includes two steps.

Inthe first step, the system performs license plate localization,which involves identifying the license plate location in theoriginal image and cropping it.

In the second step, the systemextracts and classifies all the available characters in the licenseplate, namely, numbers and letters.

2. NEURAL NETWORK BASED

For detecting the location of the license plate, artificialintelligence methods, particularly neural networks, are used.A neural network detects the location of the license plateby blurring and sweeping the image surface via a dynamicwindow. Template matching is another method, butit is highly static, and its performance is low when the exactlicense plates image, its characters, or the plate signature isapplied as a template. There are other methodsand algorithms that use plate signature, fuzzy logic, geometric parameters, and window movement. Morphological functions also give acceptable results.

3. CROPPING CAR PLATE

The license plate’s aspect ratio is standardand constant, and it can be determined by applying a verticalor a slope search. Although the length and the direction of alicense plate are available from the vertical or the slope sweep,the width of the Iranian license plate can be obtained by asimple multiple usingwhere License Plate.

Lengthis the width of the license plate, andLicensePlateis its length. Cropping begins from the top leftcoordinate of Colonyyand in the same orientation as that ofthe blue rectangle width. Cropping inthe direction of the blue rectangle causes license plate imagesthat are straight and without any angles the plate looks after cropping.

4. CONVERT TO BINARY IMAGES

A threshold value is required for converting the license plateinto a binary image. An incorrect high or low threshold causesconnectivity or discontinuity between license plate characters;hence, findings a tradeoff threshold value is important. Thisis generally done by using a histogram to find entropies andthe probability of distribution between objects/foregrounds andbackground.

5. REMOVING SURPLUS COMPONENTS

For correct recognition, the binary license plate should befree from noisy dots and surplus objects. In these step, theproposed system uses morphological functions (i.e., erosionand dilation) to delete small noisy dots and to correct edges.

FUZZY – BASED ALGORITHM

In fuzzy sets, they try to localize a license plate converted all image pixels to anothercolor space such as HSI and used tiled histogram to find backgroundand foreground colors to localize license plate texture.These are used color edge detection,such as black–white, red–white, and green–white, for localization and used color edge and the set of hue,saturation, and intensity to make an image fuzzy map for licenseplate localization. Usually, the methods for color analysis aretime consuming or involve elaborate processing.

SOFTWARE REQUIREMENTS

HARDWARE REQUIREMENTS

  • System: Pentium IV 2.4 GHz.
  • Hard Disk: 80 GB.
  • Monitor: 15 VGA Color.
  • Mouse: Logitech.
  • Ram: 512 MB.

SOFTWARE REQUIREMENTS

  • Operating system : Windows 8 (32-Bit)
  • Front End:Visual Studio 2010
  • Coding Language: C#.NET