A Sliding and Classifying Approach Towards Real Time Persian License Plate Recognition

Automatic License Plate Recognition, ALPR, is an important part of today’s traffic monitoring and toll-gate systems. It usually consists of two major parts: plate localization and character recognition. In this paper, we propose a real-time algorithm for detection and recognition of Persian license plates in four styles. To be real time, we employ simple and effective techniques and implement our algorithm in pure C++. Unlike conventional methods for finding the location of the plate based on structural features, we use a sliding and classifying approach combined with some statistical information. In recognition phase, two fast and accurate features are trained by a neural network. The proposed system is evaluated on 100 images of Iranian vehicles, taken from different highway/toll-gate cameras. In localization phase, system detects 100% of the plates properly and in recognition phase, 97.8% of characters are correctly recognized. The overall processing time for single plate is 0.06 s at resolution of 407x309 and 0.24 s at resolution of 1150×650. These specifications made our algorithm industryready and currently. It is used by several corporations working on parking management and law enforcement systems.