Moving license plate segmentation by region and corner-based approach

Abstract

A novel approach for license plate detection of vehicles in movement into dynamic scenes is proposed in this work. By mixing component-based and corner-based algorithms, this approach combines characteristics of the previous ones to improve success rate and robustness. The approach was tested on a database of 500 images taken during the day. A digital camera was mounted in front of a car and photographs of moving vehicles were captured in traffic. Success rates of 91.8%, 90.8% and 95.6% were accomplished for the three approaches despite diverse illumination and the physical conditions of the plates.