An Efficient License Plate Text Extraction Technique

Abstract

In vehicle recognition, a machine-based system is required to recognize vehicles. To accomplish this, extraction of character regions are required to perform with high degree of accuracy in real-life images datasets. In this letter, we present an innovative framework for character regions extraction from the license plate of vehicles. Our framework includes needful preprocessing steps to remove noise in images. The necessary checks are applied to structural properties of regions to result in character regions as an outcome. The proposed algorithms of character region extraction work with intra-regions and inter regions-dependent structural properties of characters. Our developed framework achieves results at par with respect to results reported in the literature. The character regions extraction and recognition results are 97 and 95%, respectively. These results are for images where all the regions are correctly extracted and recognized.