摘要
根据汽车牌照字符与背景的特点 ,提出了一种自适应的二值化方法 ,这种方法首先把图像分割成块 ,进而根据每小块中的前景的灰度更加接近小块中灰度最大值的特点 ,选取一个比较高的域值 ,然后根据原图像的灰度值对得到的图像进行区域增长 ,最后得到的二值图像既可以完整再现字符 ,同时可以消除背景中亮度较高的点 ;另外讨论了结合汽车牌照的几何尺寸的字符分割办法 ,详细阐述了横向穿越变化数和字符的横竖编码这两个新的特征的提取过程 ,最后给出了一种新的多级分类的规则 ,讨论了这种多级分类的实现算法 ,实验表明这是一种行之有效的分类方法。
An adaptive threshold method is proposed, which is depending on features of relation between the character and its background. At first, the image of the plate should be partitioned into 7×2 parts, and threshold should be determined near to the data (foreground) peak from the gray-level histogram of each sector, and the sectors of which histogram is unimodal should be ignored. And then, region growing is implemented based on the result image of the first step and the original image, and the threshold should be little smaller than the midway between the background peak and the data peak. The characters as the result of this method are integrated, and most of noisy spots in the background are removed. The characters segmentation method relayed on the number of characters of general plates and its length is discussed in this paper too. Two new features used for creating eigenspace are discussed in details. At last, a new multilevel classifying rule is given. Indicated by experiment, the classifying method is encouraging.
出处
《计算机测量与控制》
CSCD
2003年第8期574-576,共3页
Computer Measurement &Control
基金
国家 8 6 3光电子主题 30 7资助项目 [86 3- 30 7- 0 7- 0 2 (5 ) ]
