Real time license saudi plate recognition using raspberry pi

Abstract

With the advent of powerful and low cost Single Board Computers (SoC), developing fully automated systems is becoming a trivial task. We can sense objects and capture images by using embedded systems to process and perform action accordingly. License Plate (LP) recognition systems use optical character recognition techniques to detect number plate on vehicles. These systems are important to control traffic, speed, theft, toll payment and parking lot access for vehicles. In this paper we proposed Arabic LP recognition system which is applicable to the Saudi Arabian LPs with two different formats by using Raspberry Pi board. Therefore, this system recognizes both Arabic and Latin numerals and alphabets. The LP recognition process consists on many preprocessing steps to segment numbers and characters from the captured image of vehicle. The features are extracted based on the horizontal projection profiles and pixel distributions in divided zones of the characters segmented image. Further characters recognition is done for each region of the LP which allows to use SoC with limited processing power. Neural Network and distance classifier are used to recognize different characters in LP. The proposed system correctly recognizes LP with 90.6% accuracy. The low cost fully automated system is easily portable which can be deployed at remote location to control traffic.