Volume 7, Issue 5, May 2017

Volume 7, Issue 5, May 2017 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Localization Techniques in ANPR Systems: A-State-of-Art Neha Rana 1, Pawan Kumar Dahiya 2 M. Tech ECE, DCRUST, Murthal, Haryana, India 1 Assistant Professor, Dept. of ECE, DCRUST, Murthal, Haryana, India 2 DOI: 10.23956/ijarcsse/SV7I5/0338 Abstract Automated License Plate Recognition (ALPR) is used to detect, recognize a vehicle license plate. ALPR comprises of three stages namely localization, segmentation, and character recognition. This paper discusses various localization techniques and compare their performance on similar parameters. License plate Localization using signature analysis provides additional accuracy of 2% respect to other techniques stated in this paper. Keywords Automatic License Plate Character Recognition (ALPR) I. INTRODUCTION The application of computer image processing increased efficiency and security in different image analysis applications. Automatic license plate character recognition (ALPR) is one of them. ALPR system is used to detect, recognize a vehicle license plate. ALPR systems have been used in parking lots, security control of restricted areas and traffic surveillance. Vehicles have unique license plate to be recognized [9]. Therefore, image processing based algorithm is suitable to develop a vehicle license plate recognition system. The aim of this thesis is to compare various localization techniques. ALPR comprises of three stages namely localization, segmentation and character recognition. In this paper, various localization techniques are compared. Localization is a process of identifying license plate from the image captured. License plate is characterized by its dimensions and high contrast. The contribution of this is as under: The various localization techniques used in ALPR system are identified. The performance of the various techniques is compared based on similar parameters. The rest of the paper is organized as under: Section 2: Automatic License Plate Recognition system Section 3: Techniques of license plate localization Section 4: Discussion Section 5: Conclusion II. AUTOMATIC LICENSE PLATE RECOGNITION SYSTEM ALPR is a system that captures the image of moving vehicle and stores the images, localize license plate and then read characters on the plate using OCR (optical character recognition). ALPR system comprises of things stated as under: Hardware to capture image. Image processing to detect, normalize and enhance image of plate. OCR to extract the characters from the processed image. Database and standard computer hardware to recognize vehicles. ALPR works in two modes, either in real time mode or send the image to a remote location and then do the image processing. The limitations of ALPR are stated as under: Higher relative speed of vehicle to camera leads to motion blur. Low contrast and different illumination. Skew in the image captured. III. LICENCE PLATE LOCALIZATION Many different features are utilized to find the location of the license plates, e. g. shape, color, orientation and frequency. Colors and shapes of the objects are among the most popular features that are utilized [7, 11, 10]. Some systems [9] change the color space to find more robust representations. There are various algorithms used for number plate localization. Some of them are listed and described in this paper 1. Localization using Signature Analysis 2. Localization using characteristics of Alphanumeric characters 3. Localization using novel approach 4. Localization using genetic algorithm including color feature extraction 5. Localization using morphological methods 2017, IJARCSSE All Rights Reserved Page 682

3.1 Localization using Signature Analysis A license plate consists of characters that are recognized by their distinctive intensities from the background. This vertical stroke is defined as the signature of a license plate. Plate is situated in the lower part of the vehicle hence this analysis will search in the five segments for signature of a plate [2]. Each row will be divided in three rows. The program requires at least two rows to have a valid signature before the particular segment qualified for the next process. Signature is considered as repetition of vertical edges and gap between two peaks. A threshold method is used to validate a signature in a plate [2]. Adaptive thresholding is used to increase the efficiency. The Fig.2(a) shows an invalid signature for a vehicle s non-text area having gap greater than 20 pixels. While on the contrary, Fig.2(b) shows a valid signature with both number of peaks and gap between consecutive peaks fall in the range of the set threshold. Fig.2(c) shows the signature searching procedure below the license plate area with no signature found [2]. 3.2 Localization using characteristics of Alphanumeric characters Alphanumeric characters have specific properties in binary image such as size, white pixel density, atomicity etc. The steps of proposed approach for localization of license plate from an image are explained below. (A)Inverted L : Masking mask having size equal to maximum character size is used, having shape of inverted L. At each position, these conditions are tested: a) All pixels in first row and first column of a box should be black. b) There should be at least a single white pixel on second row and second column of a box. [8] Input image of the number plate is pre-processed for the removal of background noise [8]. The pre-processed image is binarized to make it suitable for localizing the number plate. Otsu s algorithm [5] is employed to for fast binarization. Fig. 1: Evaluation of Signature Analysis [2] Fig. 2: Input image [8] Fig. 3: Output of Inverted L masking [8] 2017, IJARCSSE All Rights Reserved Page 683

(C)White Pixel Density: At each pixel, white pixel density is calculated and if it lies in certain percentage of total area occupied, it is identified as a character location [8]. Fig. 4: Output of this step [8] 3.3 Localization using novel approach The proposed algorithm generally is divided into three main parts, including the license plate localization, character segmentation and character recognition. The localized plate is first part. This part nominally includes the following steps: 3.3.1 Noise Intensity is uniformed and noise is reduced using Gaussian filter approach. Intensity range is modified for noise reduction. Fig. 4: Block diagram of novel approach [6] 3.3.2 Changing color space Next, the RGB color space is transformed into gray space using the conventional formula [6]. G(r) = 0.2989 * Red + 0.5870 * Green + 0.1140 * Blue [6] 3.3.3 Intensity modification Histogram equalization technique is used to increase the intensity range. 3.3.4 Edge detection The vertical Sobel operator or Prewitt is utilized for edge detection [6]. 3.3.5 Separating objects from context Pixel connectivity are used to identify object from context [6]. 3.3.6 Finding connected component The connected objects are investigated with using 8 and 4-ary connectivity. 3.3.7 Candidate selection Candidate for license plate is selected based on area, range and ratio of length to width, and region intensities [6]. 3.4. Localization using genetic algorithm including color feature extraction Genetic algorithm identifies license plate by analysis of geometrical relationships between symbols in a plate area. 1. Image Processing Phase By image processing we mean color feature extraction, size filtering, connected component analysis and color to binary image conversion. 2. Genetic Algorithm Phase It involves initial population setup, fitness function formulation, selection method, mutation and crossover operator design and parameters setting [1]. 2017, IJARCSSE All Rights Reserved Page 684

1) Chromosome Encoding: To recognize license number, symbols including the plate number are included [1]. 2) Fitness Function: It is selected as the inverse of objective distance between two chromosomes [1]. R X1 = (X2 X1) / H RX2 = (Y2 Y1) / H 3) Selection Method: the stochastic universal sampling (SUS) method involves mapping of each individual to a continuous segment of line equal in size to its fitness [1]. 4) Mutation Operator: It is done to remove less fit members of population. 5) Crossover Operator: There are many methods used to implement the crossover operator. For instance, single-point crossover, two-point crossover, n-point crossover, uniform crossover, three-parent crossover, alternating crossover and so on [1]. 6) Elimination: Comparing with the intensity of adjacent genes and some threshold values are used to eliminate genes considered as part of plate. 3.5. Localization using morphological methods The proposed algorithm in this paper consists of three major stages: Morphological operations for extracting plate features; Selection of candidate regions; and [4] Validation of plate region. Fig. 5: Block diagram of the overall proposed system [4] Original RGB Image Plate Feature Extraction Selection Of Candidates Number Plate Validation of Plate Region Selection of Candidates Plate Region and Validation of Plate Region: The output image from the previous stage consists of a set of groups of connected pixels. Labelling is done using pixel connectivity method. Some pixels are selected then using known geometrical conditions [4]. A validation process needs to be used to extract the candidate that represents the license plate. Once acquiring the plate region coordinate, the final number plate can be extracted from the original binary image. IV. DISCUSSION Various researches have been developed using different methods to process License Plate Localization, which helps the researchers to efficiently localize the License Plates and thus lead them to the reduction of false localization rate. This comprehensive review work also provides a comparison study of each License Plate Localization method. Comparison of various techniques is done below. Technique Approach Accuracy achieved Using Signature Signature analysis, CCA, Euclidean 92% Analysis distance transform Using alphanumeric Pixel density evaluation, Filtering 90% characters Novel approach Pixel connectivity, Hough transform, Morphological operator, Skeletonizing 91% Using Genetic CCA, Genetic crossover operators, 90% algorithm Color feature Using morphological operators Morphological operation with different structuring elements V. CONCLUSION The paper discusses various localization techniques and their performance. Signature analysis is used to localize with 92% successful rate and failure due to the improper illumination and blurring. Alphanumeric characters are used to localize plate with an accuracy of 90%. Restrictions on the technique is due to variations in size and characters of number plate, font used and background and foreground colors. The novel approach localizes license plate with accuracy of 91%. Genetic algorithm, based on Darwin s theory of survival of the fittest localizes with an accuracy of 90%. The system is implemented using MATLAB and tested on various samples. Morphological operators localize with an accuracy of 91%. The method is used on database of UK number plates. 2017, IJARCSSE All Rights Reserved Page 685 91%

REFERENCES [1] Arya P. Unnikrishnana*, Roshini Romeo, Fabeela Ali Rawther, License Plate Localization Using Genetic Algorithm Including Color Feature Extraction, in International Conference On Emerging Trends In Engineering, Science And Technology (ICETEST ), 2015. [2] Lorita Angeline1, Hou Pin Yoong1, Hui Keng Lau1, Ismail Saad1 And Kenneth Tze Kin Teo2, Automatic License Plate Localisation And Identification Via Signature Analysis, Lorita Angeline1, ICTACT Journal on Image and Video Processing, Volume: 04, Issue: 03, 2014. [3] Mohamad H. Dashtban, Zahra Dashtban, Hassan Bevrani, A Novel Approach for Vehicle License Plate Localization and Recognition, in International Journal of Computer Applications (0975 8887) Volume 26 No.11, 2011. [4] Xiaojun Zhai, Faycal Benssali and Soodamani Ramalingam, License Plate Localisation Based on Morphological Operations, in 11th Int. Conf. Control, Automation, Robotics and Vision Singapore, 7-10th December 2010. [5] Zhang, Yungang, and Changshui Zhang. "A new algorithm for character segmentation of license plate." Intelligent Vehicles Symposium, 2003. Proceedings. IEEE. IEEE, 2003. [6] T. Naito, T. Tsukada, K. Yamada, K. Kozuka, And S. Yamamoto, Robust License-Plate Recognition Method for Passing Vehicles Under Outside Environment. Vehicular Technology, IEEE Transactions 2000. [7] T. Sirithinaphong And K. Chamnongthai, The Recognition of Car License Plate For Automatic Parking System, in Signal Processing And Its Applications, ISSAP 1999, proceedings of The Fifth International Symposium, Volume 1, Pages 455 457,1999. [8] Pushpendra Singh Chouhan, Localization of License Plate Using Characteristics of Alphanumeric Characters, in Proc. IEEE Lnt. Workshop on Cellular Neural Networks and Their Applications, Pp. 212-217, 1998. [9] Sang Kyoon Kim, Dae Wook Kim, And Hang Joon Kim, A Recognition Of Vehicle License Plate Using A Genetic Algorithm Based Segmentation in International Conference, Volume 1, Pages 661 664,1996. [10] J. A G Nijhuis, M.H. Ter Brugge, K. A. Helmholt, J. P W Pluim, Spaanenburg, R. S. Venema, And M.A. Westenberg, Car License Plate Recognition With Neural Networks And Fuzzy Logic in Neural Network 1995 Proceedings, IEEE International Conference, Volume 5, Pages 2232 2236, 1995. [11] Eun Ryung Lee, Pyeoung Kee Kim, And Hang Joon Kim, Automatic Recognition of A Car License Plate Using Color Image Processing, in ICIP 1994, IEEE International Conference, Volume 2, Pages 301 305,1994. 2017, IJARCSSE All Rights Reserved Page 686