Detecting, tracking and recognizing license plates

Abstract

This paper introduces a novel real-time framework which enables detection, tracking and recognition of license plates from video sequences. An efficient algorithm based on analysis of Maximally Stable Extremal Region (MSER) detection results allows localization of international license plates in single images without the need of any learning scheme. After a one-time detection of a plate it is robustly tracked through the sequence by applying a modified version of the MSER tracking framework which provides accurate localization results and additionally segmentations of the individual characters. Therefore, tracking and character segmentation is handled simultaneously. Finally, support vector machines are used to recognize the characters on the plate. An experimental evaluation shows the high accuracy and efficiency of the detection and tracking algorithm. Furthermore, promising results on a challenging data set are presented and the significant improvement of the recognition rate due to the robust tracking scheme is proved. © Springer-Verlag Berlin Heidelberg 2007.