Paper title

“A New Approach for Recognizing Saudi Arabian License Plates using Neural Networks”       

Authors: Mohamed Deriche and Muhammad Moinuddin
Affiliation: King Fahd University of Petroleum & Minerals, Saudi Arabia, Dhahran

Abstract—In this paper, a neural networks (NN) based automatic license plate recognition system (ALPR) is proposed for Saudi Arabian license plates with Arabic characters. The license plate region is [1]rst localized by [1]rst tracing the exterior and the

interior close boundaries of objects in the car image and then separating the license plate by determining the rectangularity characteristic of these close objects. Character segmentation is performed via vertical and horizontal projection pro[1]les. Finally, a Multilayer Feedforward Neural Network (MFNN) with a backpropagation (BP) algorithm is used for character recognition. We discuss new features from the characters for training the NN. The results obtained from a medium size data base are very promising, i.e., 98%. The alogoritms discussed here were tested at the entrance of a praking lot to mimic a real life situation.