A Real Time Vehicle’s License Plate Recognition System Choudhury A. Rahman, Wael Badawy Department of Electrical and Computer Engineering University of Calgary Calgary, Alberta, Canada T2N 1N4 {rahmanc, badawy}@enel.ucalgary.ca Ahmad Radmanesh Signals Division The City of Calgary Calgary, Alberta, Canada T2P 2M5 aradmanesh@calgary.ca Abstract A smart and simple algorithm is presented in this paper for vehicle’s license plate recognition system. Based on pattern matching, this algorithm can be applied for real time detection of license plates for collecting data for surveying or for some application specific purposes. The proposed system has been prototyped using C++ and the experimental results have been shown for recognition of Alberta license plates. 1. Introduction Vehicle’s license plate recognition system has been a special area of interest in video surveillance arena for more than a decade or so. With the advent of sophisticated video vehicle detection systems for traffic management applications, number plate recognition system finds wide varieties of places to fit itself beyond just controlling access to a toll collection point or parking lot. It can now be integrated to the video vehicle detection systems which usually are installed in places of interest for intersection control, traffic monitoring etc., to identify vehicle that violates traffic laws or to find stolen vehicles. There are a number of techniques used so far for recognition of number plates such as BAM (Bi-directional Associative Memories) neural network character recognition [1], pattern matching [2] etc. The technique used in the proposed system is based on pattern matching, which is fast and accurate enough for real time applications and is developed for recognition of Alberta license plates with prior knowledge of letters and numbers orientation. Since the orientation and font used for number plates differ in different countries/states/provinces, this algorithm is needed to be modified accordingly keeping its structure intact, if we want to apply this system for recognizing the number plates of those places. The rest of the paper is organized as follows: section 2 discusses about the system architecture and the steps involved in recognition process, section 3 explains the proposed algorithm for real time detection of number plates, section 4 shows the experimental results and finally section 5 concludes the paper with acknowledgements and references. 2. System Architecture The proposed system will be used as a part of intersection surveillance video camera system for traffic analysis. Figure 1 shows a typical intersection of the city of Calgary. Only one license plate is used in Alberta province which is attached to the back side of the vehicle and the camera will be used to track this back license plate. Figure 1. A typical intersection of the city of Calgary The proposed system architecture contains three distinct parts: outdoor part, indoor part and communication link. Outdoor part is the cameras installed in different intersections of interest for capturing images. The indoor part is the central control station that receives, stores and analyzes the captured images from all these installed cameras. Communication link can be high speed cable or fiber optic connecting all these cameras to the central control station. Almost all the algorithms developed so far work by following similar steps. Seven general processing steps have been identified as being common to all number plate recognition algorithms [3]. These are: Trigger: This may be hardware or software trigger. Hardware trigger is the old approach where inductive loop is used for triggering and this tells when the image should Proceedings of the IEEE Conference on Advanced Video and Signal Based Surveillance (AVSS’03) 0-7695-1971 3 $17.00 © 2003 IEEE