摘要
针对在复杂背景图像中汽车牌照检测定位的课题,提出了一种基于可信度评价的汽车牌照检测定位算法。在对图像进行预处理后,算法利用车牌区域由于字符排列的规则性而在梯度图像上产生的特征,分割并提取出车牌可能存在的候选区域,然后根据汽车牌照的几何特征对每一个候选区域进行可信度评价并按一定的合并规则对候选区域进行合并,以获取更高的可信度,最后根据区域可信度值的大小确定车牌的位置。算法对于图像光照条件变化、视角变化而造成的车牌的倾斜和变形等情况都具有较好的处理效果。
The issue of a novel method to extract car license plates from a complex scene is presented by considering both the distributive regulation of the characters in a license plate and the geometrical features of a license plate. A segmenting algorithm is used to look for candidate regions that probably contain characters in a range of sizes. Each candidate region is given a credit value to measure its likelihood to be a license plate and these regions are combined according to some rules to get a higher credit value. The car license plate can be found according to its credit value. Experimental results show that the algorithm is robust in dealing with different conditions such as poor illumination and distortion of a license plate generated by different visual angles.
出处
《武汉理工大学学报(信息与管理工程版)》
CAS
2005年第3期37-40,共4页
Journal of Wuhan University of Technology:Information & Management Engineering
关键词
汽车牌照检测
直方图均衡化
均值滤波
可信度评价
license plates detection
histogram equalization
mean filter
credit evaluation