License plate location based on combinatorial feature

Abstract

The license plate location is an essential process in the license plate recognition system, and has been studied widely. In view of vehicles image on different background and light densities, this paper presents a license plate location algorithm based on combinatorial feature. Firstly, we propose to preprocess the image of license plate with traditional but rapid method. Secondly, we define some useful features so as to roughly locate the license plate. At last, we repeatedly use the combinatorial feature to obtain the real license plate from many candidates. We design this novel algorithm and integrate it into the project of the license plate recognition system. We implement and evaluate our system in various illumination and background conditions. Through extensive experiments, we show the algorithm of this paper can extract the license plate from the image efficiently and accurately.