A novel approach for license plate localization based on SVM classifier

In this paper, we present an accurate and robust license plate localization approach based on the Support Vector Machine (SVM) classifier. We impose the license plate localization as a classifier based binary recognition problem. The images of our database exhibit various illumination conditions (daytime, dark night). For the problem of illumination effects, we use the algorithm of Scale Invariant Feature Transformation (SIFT), because the advantages of SIFT feature are scale-invariant, luminance-invariant. In this way, we can locate the license plate efficiently for the vehicle image which shot at night. Experimental results illustrate the great robustness and efficiency of our approach.