基于Adaboost的车牌定位系统设计与实现-软件工程专业论文.docxVIP

AbstractThis Abstract This paper discusses the method and process of the design and implementation of the vehicle license plate localization system，vehicle license plate localization system is the first part of the license plate recognition system，the other two parts are license plate segmentation，license plate recognition．It is directly related to the recognition rate of the whole system，and it has a great impact on the recognition speed． This paper includes three parts on theoretical research，they are feature extraction， selection of algorithms and improved algorithm．By fully studying Haar features，we proposed two types of Haar-like feature．They are‘hui’’glyph feature and asymmetric linear feature．We use the original Haar feature and two new Haar-like features to be the feature template，and optimize the window’S size and step to generate 5571 features and weak classifiers．Compare to the Haar-only system，it improves the accuracy of the vehicle license plate locMization system．In the implementation of the AdaBoost algorithm，on the one hand，with the help of cascade strong classifier, recognition rate is much beRer than the original strong classifier．On the other hand， we improve the AdaBoost algorithm by cache the value of features before iteration of the training process，which greatly improves the performance of the training algorithm． Finally,we C++programming language to develop this license plate localization system in Microsoft Visual Studio 2008 programming environment．1 50 images are used to be the test set，and the recognition rate is 9 1％．The experimental results show that the license plate localization system based on Haar features， Haar-like features and AdaBoost algorithm can do a good job on real—time license plate location．It has high recognition rate and low false detection rate． Key Words：License Plate Localization；PaRem Recognition；AdaBoost II 万方数据 目 目 录 第一章绪论 ．1 1．1研究目的及意义 。1 1．2国内外发展现状 。1 1．3系统应用前景 3 1．4论文研究内容 3 1．5论文组织结构 5 第二章基本概念及相关技术介绍 6 2。1特征