Character and Font Recognition in Hybrid Graphic-Text Plate Numbers

Abstract

The varied plate designs and character fonts of Philippine licence plates make it difficult to read the characters. This study focuses on enhancing the segmentation approach for character recognition on various licence plate types used in the Philippines. The suggested system includes character segmentation, character recognition, and licence plate categorization. The colour level of the image's pixels was used to categorise different licence plate series. Three-Class Fuzzy Clustering with Thresholding and Connected Component Analysis were used to segment plate characters, and Template Matching was used to identify them. After testing 20 licence plates from each series, the system's accuracy for the 2003 plate series and 2014 plate series, respectively, was 95% and 70%.