【作者】 尹令；

【导师】 黄辉先；

【作者基本信息】 湘潭大学 ， 电力电子与电力传动， 2003， 硕士

【摘要】 本文对车牌自动识别系统中图像预处理、车牌定位分割、字符切分和字符识别等模块涉及的新算法、新技术以及系统设计进行了比较全面的论述。针对目前的研究情况，对一些关键技术进行深入研究，主要解决下面几个方面的问题：(1)如何准确快速的定位分割牌照；(2)如何精确的分割出每个字符图像；(3)怎样设计高识别率的识别决策；(4)规划系统设计和需要实现的功能。 在车牌定位分割研究方面本文提出一种线段性探针搜索定位分割算法。该算法充分利用车牌固有特征采用线段型结构元素逐行扫描图像，统计与之匹配的结构象素点进行分析，能准确而快速地检出车牌区域。关于字符切分，本文研究并提出基于垂直投影差分分析的切分算法。这种算法的特点是研究对象为垂直投影值经过阈值筛选后再做前向差分运算的差分结果。这个差分值反映了垂直投影图的峰谷宽度信息。针对车牌大小不一，存在少量污渍和其他干扰的车牌有较好的分割效果。字符识别模块，分析模板匹配和神经网络的优势，采取了基于字符外轮廓特征匹配和神经网络组合识别的方法。这样有利于吸取两种方法的优势，弥补不足。最后讨论如何在VC++6.0中设计和实现车牌自动识别系统，并对处理后的结果进行分析，对论文研究工作进行总结和展望。更多还原

【Abstract】 In this article, some new algorithms and methods are presented in such stages as pre-processing of images, license plate segmentation, character segmentation, character recognition in the license plate recognition system. At the same time, some crucial problems of technology were overall discussed, for example, (1) How to locate and segment license plate rapidly and exactly? (2) How to segment each character image accurately? (3) How to design the high recognition ratio decision? (4) How to design and implement the system?This paper presents an approach called linear probe search to locate and segment license plate, which makes use of the characteristic of the license plate using a linear structure to scan the image and record the marching pixels. About the character segmentation, this article studies the method of based on vertical projection difference analysis, which can reflect the width of apex and paddy. In the recognition of characters module we present an approach of based on characteristic template matching combined with neural network recognition. Finally we specify the design and realization of the license plate recognition system on the plat of VC++6.0.更多还原