车牌字符识别系统的研究与实现-控制工程专业论文.docx

华中科技大学硕士学位论文 华 中 科 技 大 学 硕 士 学 位 论 文 II II Abstract License plate recognition (LPR) system can identify the license plate accurately and efficiently. The LPR system can greatly reduce the labor of mankind and bring economic benefits. Therefore, LPR system has a very large application prospect. LPR system is comprised of two major phases: license plate character location and character recognition. In the LPR system, the location of the plate area is detected firstly. The coordinate information of every character in the plate can be obtained by taking the plate location rules into consideration. The entire plate is identified by the recognition of every character. In The character location module, we use the edge point density to get the location of the candidate plates. Then purpose the methods of color pixel distribution to wipe off some false plates. At last use the method of image segmentation and connected region projection segmentation to obtain the coordinate information of every single character. The algorithm purposed in the character recognition module is based on the random forest classification technique and the Adaboost cascade classifier technique. In order to achieve an accurate and efficiently recognition result, we use the Adaboost cascade classifier to train the similar characters and characters difficult to be identified individually. In the experiment, we use 1000 positive and negative samples to train an offline classifier. The experimental results based on 1000 images different from the training samples shows that the algorithm’s accurate result is 87.5%. Keywords: License plate recognition Character recognition Random forests Adaboost cascade classifier PAGE IV PAGE IV 目 录 摘 要 I HYPERLINK \l _bookmark0 Abstract II 1 绪论 HYPERLINK \l _bookmark1 1.1 研究背景 (1) HYPERLINK \l _bookmark2 1.2 研究现状 (2) HYPERLINK \l _bookmark3 1.3 车牌识别的主要流程 (5) HYPERLINK \l _bookmark4 1.4 论文的研究内容和结构 (6) 2 车牌字符定位 HYPERLINK \l _bookmark5 2.1 基于边缘密度的车牌区域定位方法 (8) HY