Adaptive character recognition using evolving templates

Abstract

Optical Character Recognition is widely used in today's world. It can be used in a variety of applications and environments such as offices, libraries or even at home. In this project Optical Character Recognition techniques are applied to Automatic Number Plate Recognition environments, which involves the scanning and interpreting of Vehicle License Plates. Template-matching techniques are used to interpret characters detected and extracted from an image, specifically Vehicle License Plates. The aim of this project is to create the best representation from among the characters extracted for each specific character. This being an NP hard problem, Genetic Algorithms are employed so as to reach a solution in a reasonable time. Another aim of this project is to perform a reasonably high accuracy rate, using template-matching, and this is achieved. Other techniques such as Connected Components Analysis and thresholding are used to extract the data from the images. In the end, the Genetic Algorithm performs its job well, and manages to create a set of characters, representative of the characters previously encountered.