International Journal of Scientific and Research Publications, Volume 4, Issue 9, September 2014 1 ISSN 2250-3153 www.ijsrp.org Image Acquisition Image Pre- Processing Character Segmentation Block Based Feature Extraction Recognition Block-Based Neural Network for Automatic Number Plate Recognition Deepti Sagar*, Maitreyee Dutta** * M.E, Computer Science, NITTTR, Chandigarh **Professor & Head, Department of Computer Science, NITTTR, Chandigarh. Abstract - Automatic Number Plate Recognition (ANPR) system is highly accuracy-demanded application for car identification. This paper presents a new method of block-based ANPR system for recognition of Indian car license number plates. Since number plate guidelines are not strictly practiced in India and wide variations found on these plates in terms of font type, character size, screws/dots etc., it often becomes difficult to correctly identify the non-standard number plate characters. This proposed method works well with both standard and non- standard Indian number plate images taken under various illumination conditions. By using the block-based feature extraction process this method of recognition provides a high recognition rate of 98.2% and speed up the processing time of each character to 3.3ms when using a database of 3399 character images. Index Terms:Block-Based Character Recognition, Image acquisition, Image enhancement, Neural Networks,Segmentation. I. INTRODUCTION Automatic Number Plate Recognition (ANPR) system is an image processing system, which lies under the computer vision field. It has been a special area of interest due to its many applications such as for traffic law enforcement; find stolen cars, parking lots and surveillances [4]. ANPR is used to identify vehicles by capturing license plates and recognize the characters. The software of recognition process generally composed of four main stages: 1) Image enhancement, 2) Segmentation, 3) Feature extraction and 4) Character recognition. This paper will discuss these stages in detail. A wide variety of techniques have been developed in the past, but most of them worked under restricted conditions and causes challenges in recognition task such as, projections and pixel connectivity are the most common methods for segmentation [1], [6], [7], [8]. There are also some paper proposed segmentation methods are using prior knowledge of characters [4], [12], character contour [14], combined features [11]. For the recognition of the characters, many classifiers can be used such as the most common used Artificial Neural Networks (ANN) is feed-forward ANN which has a simple architecture as compared to the other common pattern matching techniques like Self-Organizing neural network having problem with joined and missed characters, template matching which can recognize only single font, fixed size characters [1], [4], [9], [11]. Other methods like Normalized Cross-Correlation (NCC) and Support Vector Machine (SVM) having high computational cost, HNN requires too much memory and fuzzy logic does not work well with bad quality images [13], [3], [2], [4]. The current methods of ANPR system worked accordingly to the guiding parameters of specific country traffic norms and standards [5]. Although, in India, number plate standards exists, but they are rarely practiced. As a result, wide variations are found in the number plates, in terms of font type, character size, screws/dots and location of the number plate, also many unnecessary characters are present on the number plate. Various other issues involved in the number plate recognition in terms of plate and environmental variations. The aim of this study is to develop a Block-Based ANPR system for recognition of Indian car license number plates by resolving these issues with non-standard number plates, to provide high recognition rate and to speed up the processing time as compared to the other ANPR system based on neural network in [13]. The proposed algorithm has been implemented and tested with a database of 3399 Indian binary character images using MATLAB. The rest of this paper is organized as follows: Section II describes the proposed methodology used to develop an ANPR system. The MATLAB implementation and analysis of the results are presented in Section III. Section IV concludes the paper. II. METHODOLOGY The proposed Block-Based recognition system using neural network introduce a new method for segmentation and feature extraction process to extract the character features, which have a great effect on recognition process. By optimizing these two steps before recognition, the proposed system gives good results of recognition using feed-forward Artificial Neural Network. The proposed approach, use these basic concepts for each module as shown in the Figure 1: image pre-processing system and projection profiles for segmentation, block-based feature extraction using edge density calculations and neural network for recognition. Figure 1: Modules of the Proposed System