Automatic License Plate Recognition from Live streaming using Stroke width Transform and Artificial Neural Network

Abstract

As the public and private transportation system increases, there is an urgent need of an Automatic License Plate Recognition (ALPR) system. ALPR system is equipped with many intelligent surveillance systems like road traffic management, security management and automatic toll collection system, etc. The basic steps in ALPR were accurate localization of number plate and recognition of license plate characters, which bears on the overall system accuracy. A license plate detection method was produced to find number plates from a live snap shots of a video stream showing the movement of all the vehicles in various conditions such as, non-uniform illumination, vehicle speed, background and foreground color, different weather condition, occlusion within image, etc. In this composition, the detection of a license plate from an image, stroke width transform was applied and been simulated on live snapshots. The artificial neural network based character and number recognition was applied on the detected license plate. The proposed algorithm gives 98% of license plate detection accuracy and 92.7% of number plate recognition accuracy.