A Thai license plate localization using SVM

This paper proposes a method for localizing a Thai license plate from an image. The proposed method contains three main processes of: 1) a pre-processing; 2) a sub-image analysis; and 3) a license plate classification. In the pre-processing, a canny edge detection is applied to convert a given image into a corresponding edge image. This process helps to reduce image's noise caused by a cluttered background of the image and a cluttered background of the license plate itself. In the sub-image analysis, a sliding window technique is used to create a region of interest (ROI) which moves in pixels along both vertical and horizontal directions of the image. Then, in the license plate classification, a support vector machine (SVM) is employed as a classification tool which is used to distinguish a license plate from other objects. The trained SVM model is applied on ROIs in order to identify the license plate. In the experiment, a dataset of Thai license plates under some difficulties e.g. view variations and cluttered backgrounds is used to validate the promising performance of the proposed method.