A license plate detection and character segmentation method under difficult conditions

Abstract

This paper presents a novel frame-by-frame license plate detection and character segmentation method for security applications. The system is based on edge statistics, mathematical morphology and typical license plate grayscale pattern. The system locates and extracts Israeli license plates from a grayscale video which was recorded under difficult illumination conditions. While existing systems usually operate in a lenient environment, the proposed system works under highly complex conditions. The system consists of three stages: 1) Extraction of the region of interest (ROI). 2) License plate search and classification of possible candidates. 3) Character segmentation. The system was tested on a video data containing 4668 frames, and achieved a success rate of 86.6%.