Abstract: License plate automatic recognition is one of the key technologies of intelligent transportation system, which contains two parts of license plate detection and character recognition. In this paper, we propose a learningbased method using coarsetofine strategy to realize fast and accurate license plate detection and recognition. In detection stage, colorpair and vertical edge features are firstly used to find the rough position of the license plate, and then a machine learning method based on HOG feature and SVM classifier is adopted to determine the accurate plate position. In recognition stage, characters are firstly separated based on connected component analysis and the inherent character features, then three kinds of stable and effective features are extracted according to character structures and SVM classifiers are applied to recognize the characters. Experimental results on a database consisting of 412 car images with different viewing angles and different illumination conditions demonstrate the effectiveness of the proposed method.