基于数学形态学的图像分割方法研究
专业:电子信息科学与技术
班级:2005级1班
姓名:杨晓琦
引言 3 1 图像分割基本理论7
1.1 图像分割的概念7 1.2 传统的图像分割方法9 1.3 特殊理论工具的图像分割方法 12
1.4 图像分割的评价 13
2 数学形态学基本理论16
2.1 形态学的概念 16 2.2 结构元素的选取 16 2.3 二值形态学理论 18 2.4 灰值形态学理论 20 2.5 形态学重建 21 2.6 形态学边缘检测 22
3 Matlab在图像分割处理中的应用 24
3.1 Matlab简介.24 3.2 Matlab在图像处理方面的应用.24
3.3 基于Matlab的图像分割.26
4 车牌图像分割的相关理论研究28
4.1 车牌定位算法简介 28
4.2 车牌的字符图像分割 37
5 基于数学形态学车牌图像分割42
5.1 形态学车牌定位 42 5.2 形态学字符图像分割 53 5.3 本章小结 61 结论62
致谢63 参考文献64附录1 源程序清单 68
本文运用形态学方法对车牌定位算法和车牌字符分割算法进行了系统的研究。
这两种算法的研究为车牌识别做了先期准备,是智能交通中非常重要的组成部分。
在车牌定位算法部分,提出了一种基于二值面积形态学的车牌定位算法。
首先将车牌的灰度图像二值化,然后逐步缩小车牌候选区的面积,计算车牌图像中连通区域的面积,并根据车牌图像的实际情况确定面积阈值,并用形态学的方法对车牌图像进行处理运算,以实现车牌的精确定位。
仿真实验结果表明此算法定位精度高,而且能适应复杂天气环境,能达到满意的定位效果。
在字符分割算法部分,对投影分割算法进行了改进,将其与数学形态学分割算法进行了结合。
首先将车牌图像二值化,然后用数学形态学分割方法结合水平与垂直投影分割方法,确定车牌字符宽和高并校正车牌实际位置,通过两次投影,校正车牌角度,去除车牌边框,确定车牌上下边界及中心点,最后分割提取车牌字符。
实验结果表明该算法能有效的保持车牌字符边缘,获得较好的分割效果。
关键词:
图像分割;数学形态学;车牌定位;字符分割; Matlab
In this paper,the license plate location algorithm and the license plate character segmentation algorithm based on morphological method are studied systematically. The research of these two algorithm is an important part of intelligence traffic and serves for the first step of license plate recognition. In license plate location, an algorithm based on two-valued area morphology is presented.Firstly,the gray image of license plate is binary.then the areas of candidate registration is narrowed gradually,the area of regional connectivity is calculated. According to the actual situation of license plate determines the area thresholding,and use morphology method processing license plate image, achieve accurate positioning. The simulation results show the algorithm has a high positioning accuracy and adapts to the complex weather and environment, can achieve satisfactory results.In character segmentation algorithm, the projection algorithm is improve and combined with mathematical morphology segmentation algorithm. Firstly the license plate is binary, then mathematical morphology segmentation algorithm is combined with horizontal and vertical projection segmentation method, to determine the license plate character width and the height, and calibrate its location through two times projection.The license plate angle is calibrated, license plate frame are removed,the upper and lower boundaries and the center of licence plate are determined.Finally, the license plate character are divisional and extracted Experimental results show the algorithm can remain the license plate character and obtain good segmentation result.
Keywords:
Image division; mathematics morphology; vehicle plate location; charactersegmentation; Matlab。