A Location Method for Multi-style License Plates Based on Character Classification

Abstract

Abstract To locate the license plate exactly is a challenging task, especially for the license plates with complex backgrounds or shadows. Firstly, we use a multilevel local adaptive thresholding method instead of global thresholding methods to produce binary images under varying illuminations. Secondly, the connected contour analysis is applied to extract the main characters, and the false character contour regions are filtered out by the SVM classification method using PCA dimensionality reduction. Finally, we merge the remaining characters on the same line to extract candidate regions of the license plate. Experimental results show that the proposed method achieves higher accuracy, more acceptable and precise results. By using the data sets of the license plates in the USA and Europe, the method is shown to be suitable for multi-style license plates.