The algorithm of automatic license plate recognition is deeply studied for the main work of this article, focusing on the license plate location, tilt correction and character segmentation algorithm. Before location, a variety of traditional methods is used to pre-process vehicle image, and got more prominent license information. For locating plate, according to texture features, rough location is conducted, getting a few candidate regions. Then using mathematical morphology methods with license plate aspect ratio for fine positioning, license plate location is better achieved. For the tilt of the plate,an improved Hough transform algorithm named the hierarchical Hough transform is presented in this paper, which can effectively improve the speed. In order to accurately separate out the license plate characters, in this paper, the character segmentation algorithm of the combination of mathematical morphology and the vertical projection is proposed, which can effectively overcome the characters in adhesion, etc. Finally, the BP artificial neural network based character recognition algorithm is briefly described, and VC + +6.0 is used for programming. |