# INTRODUCTION utomatic License Plate Recognition (ALPR) is an important area of research due to its applications. It is a machine vision technique used to identify vehicles by their license plates without direct human involvement. The Intelligent Transportation System provides the data of vehicle information which can be used in follow up, analysis and monitoring. The complexity of automatic license plate recognition work varies throughout the world. For the standard license plate, ALPR system is easier to read and recognize. In Bangladesh this task becomes much difficult due to variation in plate model and specialties of Bangla scripts. Bangla text can be partitioned into three zones. The upper zone indicate the portion above the headline called 'Marta', the middle zone indicate the portion of basic characters and compound characters under the head-line and lower zone is the portion where some of the modifiers can exist in .The ALPR algorithm consists of three steps: license plate locating as well as true license plate extraction, character segmentation and character recognition. From the input image, the license plate detection, noise removal, invert and skew detection is performed from license plate in license plate extraction phase. In character segmentation phase each and every character is isolated and segmented accordingly followings step line segmentation, word segmentation and character segmentation, based on the selection of good features of characters, each character is recognized using neural network, in the character recognition phase. Extraction of license plate is difficult job; mainly due to license plate occupy a small part of whole image, difference in license plate models and cause of environmental factors. This step affects the accuracy of character segmentation and recognition work. The rest of this paper is prepared as follows segment II shortly illustrates the applications of the ALPR system in different areas, Segment III described related works, Segment IV described true license plate extraction process, Segment V includes steps of the character segmentation and Segment VI includes steps character recognition process. Segment VII discusses its experimental results and lastly Segment VIII concludes the paper. II. # APPLICATION ALPR system is mostly used in Intelligent Transportation System. ALPR is important in the area of highway toll collection, traffic problems, borders and custom security, premises where high security is needed, like national assembly, V.I.P houses and so on. plate extraction. Hao Chen [1] et al planned the method, several candidates based on texture information similar plate detection technique using group of lines forming rectangle at the plate boundary, followed by this step is the vertical edge density technique to find out the plate area. Ozbay et al [3] developed smearing algorithm to locate the license plate. Mei Yu et al [4] proposed vertical edge detection followed by size, shape filter for edge area and edge matching technique based on plate model. Farhad Faradji et al [5] first used Sobel edge detection on the image. Next, vertical projection analysis was used to locate plate area. False candidates were removed using compact factor, which estimated the densest vertical edge area declaring true license plate. Every character on detected license plates is segmented in character segmentation step. Segmentation techniques based on projection analysis, Hough transform, region growing are proposed in the text. Xinagjian He et al [6] used horizontal and vertical projection analysis for character segmentation. Yuangang Zhang et al [7] developed character segmentation using Hough Transform. In this, horizontal edges of the plate area were decided initially, using Hough Transform, which helped to segment the characters with the large rotation. Characters were segmented using vertical projection analysis based on the prior knowledge of the plate model. Feng Yang et al [8] developed region growing algorithm for character segmentation. Shen Zheng Wang et al [9] used connected component analysis for character segmentation. # III. # RELATED WORK In this paper ALPR work for Bangladeshi car is presented. Images are taken out with different lighting conditions, different background and direction. Histogram equalization, median filter are used which take care of lighting and contrast problem. Sobel vertical edge detection and morphology is employed to locate the license plate. Horizontal and vertical scanning is used to segment the line, word and characters. For character recognition work chain code and neural network is used. IV. # LICENSE PLATE EXTRACTION License plate extraction is the key step in ALPR system, which maintains the accuracy of the system significantly. The goal of this phase, given an input image, is to produce a region that contains true license plate. # a) Image capturing and noise removing In this system a high resolution digital camera is used to capture an image. Images are taken in different background, illumination conditions and at various distances from the camera to vehicle. Images are resized to (1024 X 768). All the processing steps are executed on gray scale image. Preprocessing is mainly used to enhance the processing speed, improve the contrast of the image and to reduce the noise in the image. Therefore, it is necessary to filter this noise before we process the image. Usually the low-pass filter approach is used to reduce the problem of low quality and low contrast in vehicle images. Original image that captured by digital camera and corresponding gray scale image are showed in Fig. 1 The license plate region contains plentiful edges with respect to background. Sobel edge detection is used to find out the regions which have high pixel variance value. To extract candidate license plate area from the entire image, threshold is used to select rows which are having particular white pixel density. Fig. 2 shows the result of effect of using, Sobel edge detection and threshold [10]. Morphological operations aim to remove unrelated objects in the image. Dilation and erosion are used to extract candidate plate areas from the entire image. Sometimes background areas may also get declared as candidate plate. Hence to remove the fake candidates, plate validation is done using the aspect ratio of the plate and horizontal cuts in the license plate [11]. Invert and threshold operation is performed for true license plate extraction. The captured image may be skewed, so skew detection and correction necessary to make text lines to license plate are extracted and auto-correlation based binary image and projection algorithm are used to verify the true candidate plate. Gisu Heo [2] developed license horizontal. Skew angle makes the text lines of the document image with the horizontal direction. Skew correction can be accomplished in two steps. Firstly, we will estimate the skew angle and secondly, we will rotate the image by skew angle. # CHARACTER SEGMENTATION Character isolation from the license plate region is very important and crucial step in ALPR system, which influences the accuracy of character recognition significantly. The goal of this phase, given the license plate image, is to segment all the characters, without loosing features of the characters. This phase consists of the sequence of operations as, line segmentation, word segmentation, character segment and connected component analysis. Each line is scanned vertically for word segmentation. Number of black pixels in each column is calculated to construct column histogram. The portion of the line with continuous black pixels is considered to be a word in that line. If no black pixel is found in some vertical scan that is considered as the spacing between words. Thus different words in different lines are separated. So the image file can now be considered as a collection of words. Fig. 7 shows the word segmentation process. Detection of character between headline removing the headline the characters in a word are isolated and can easily be separated. Vertical scan is initiated from the row that is just beneath the 'Matra' row to find the differentiation line between characters. If during scan, one can reach the base line without touching any black pixel then this scan successfully found a differentiation line between characters. Fig. 9 illustrates the character segmentation process. For a closed boundary, its chain code obviously depends on the starting point of the boundary. To make it invariant to the starting point, the chain code can be normalized according to the following method [13]: A chain code can be treated as a circular sequence of direction numbers. Therefore, the starting point can be redefined so that the resulting sequence of numbers forms a minimum integer. A neural network [14] is a extremely parallel distributed processor that has a propensity for storing knowledge. First knowledge is acquired by the network through a learning process and then storing knowledge is for recognition of character. We use feed forward neural network for the classification and recognition Bangla characters. We trained the neural network by normalized feature vector obtained for each character in the training set. Our layer neural networks have been used with two hidden layers for improving the classification capability. For 32 dimensional feature vectors and 4 layers is used for each Character and recognized using stored knowledge of the network. Use of two hidden layers increases the recognition rate extensively. # CONCLUSIONS An algorithm for vehicle License plate extraction, character segmentation and recognition is offered. Experiment consists of images with different size, lighting, background, camera angle and distance etc. The experimental results show that, license plates are extracted truly with higher success. Character recognition phase using connected component analysis, Freeman chain code, and Neural network that works well. We suggest to use multilayer feed forward neural network for classify and recognition of character. Recognition performance will be increased if the network trains with distorted character and with good shaped character. ![(a) and (b).](image-2.png "") 1![Fig.1 : (a) Original image (b) Gray scale of image b) Vertical Edge detection](image-3.png "Fig. 1 :") 2![Fig.2 : (a) Sobel vertical edge detection (b) Effect of threshold c) Candidate Plate Area Detection](image-4.png "Fig. 2 :") 3![Fig.3 : (a) Extracted Candidate plate (b) After Inverting and threshold](image-5.png "Fig. 3 :") 4![Fig.4 : (a) Skewed image (b) After skew correction e) True license Plate Extraction After the detection of candidate license plate area, Bounding Box analysis is used to extract plate area from the image. From the Bounding Box analysis, respective row and column indices of plate area are found out. The result is as shown in Fig. 5.](image-6.png "Fig. 4 :") 5![Fig.5 : Extracted true license plate](image-7.png "Fig. 5 :") ![a) Line segmentation Line segmentation has been executed by scanning the input image horizontally. Frequency of black pixels in each row is counted in order to construct the row histogram. The position between two consecutive lines, where the number of pixels in a row is zero denotes a boundary between the lines. Line segmentation process shown in figure.](image-8.png "") 6![Fig.6 : Line segmentation b) Word Segmentation](image-9.png "Fig. 6 :") 7![Fig.7 : Word Segmentation c) Zones of Bangla scriptFrom fig.8we see that Bangla text may be partitioned into three zones. The upper zone denotes the portion above the headline, the middle zone covers the portion of basic characters below the head-line and lower zone is the portion where some of the modifiers can reside. The imaginary line separating middle and lower zone is called base line.](image-10.png "Fig. 7 :") 8![Fig.8 : The zones of bangle script d) Detection and Deletion of MatraTo segment the character separately from the segmented word, Firstly we find out the headline of the word which is called 'Matra'. From the word, a row histogram is constructed by counting frequency of each](image-11.png "Fig. 8 :") 9![Fig.9 : After Matra elimination](image-12.png "Fig. 9 :") 10![Fig.10 : Character segmentation e) Detection below the baseline A greedy search technique is initiated for the presence of black pixels below the baseline, the result will some connected components below the baseline. The components below the baseline contain lowest point called 'Base point'. Baseline is highest frequency row of base points. After determining the baseline, The Depth First Search (DFS) technique easily extracts the characters below the baseline.](image-13.png "Fig. 10 :") 11![Fig.11 : Below the base line detection](image-14.png "Fig. 11 :") 12![Fig.12 : Slope convention for Freeman chain code](image-15.png "Fig. 12 :") 13![Fig.13 : Bidirectional chain code](image-16.png "Fig. 13 :") 14![Fig.14 : A neural network with 4 layers](image-17.png "Fig. 14 :") ?, Sameya Akter¥2011December69A?Author ¥ : B.Sc. in engineering degree in Computer Science andDifferent techniques are developed for licenseEngineering (CSE) from Dhaka University of Engineering andTechnology (DUET), Gazipur-1700, Bangladesh.Email : sameyaakter@yahoo.com , PH : +88 01726133150.© 2011 Global Journals Inc. (US) © 2011 Global Journals Inc. (US) Global Journal of Computer Science and Technology Volume XI Issue XXI Version I DecemberAutomatic License Plate Recognition (ALPR) for Bangladeshi Vehicles * A novel method for license plate localization JishengHao Chen HuachunRen JianqunTan Wang 4th Proc. of ICIG 2007 * Extraction of car license plate regions using line grouping and edge density methods GisuHeo MinwooKim InsookJung International Symposium on Information Technology Convergence 2007 Duk Ryong Lee,Il Seok Oh * Automatic vehicle identification by plate recognition SerkanOzbay ErgunErcelebi Proc. of PWASET of PWASET 2005 9 * An approach to Korean license plate recognition based on vertical edge matching MeiYu Yong DeakKim IEEE International Conference on System, Man and Cybernetics 2000 4 * A Morphological based License Plate Location FarhadFaradji MajidAmir Hossein Rezaie Ziaratban ICIP 2007 * Segmentation of characters on car license plates XiangjianHe 10 th Workshop on Multimedia Signal Processing Oct. 2008 * A New algorithm for character segmentation of license plate YungangZhang ChangshuiZhang Proc. Of IEEE Intelligent Vehicles Symposium Of IEEE Intelligent Vehicles Symposium 2003 * A Novel approach for license plate character segmentation FengYang ZhengMa MeiXie ICIEA 2006 * Detection and recognition of license plate characters with different appearances Shen-Zheng His-JianWang Lee Proc. of 16 th International Conference on Pattern Recognition of 16 th International Conference on Pattern Recognition 2003 3 * Digital Image Processing CRafel RichardEGonzalez Woods 2002 Prentice Hall, Inc 2nd edition * Combining Hough transform and contour algorithm for detecting vehicle license plate Tran Dur Duan AnhDuong TranDuc HongLe Du Proc. Of International Of International * Symposium Intelligent Multimedia, Video and Speech Processing 2004 * Computer processing of line drawing images HFreeman Computer Survey 6 1974 * RGonzalez REWoods Digital Image Processing Prentice Hall 2002 * SHaykin Neural Networks, A Comprehensive Foundation 2nd Edition