基于投影特征和先验知识的车牌字符分割算法  

Plate Character Segmentation Algorithm Based on Projection Feature and Prior Knowledge

作　　者：刘军[1] 向军[1] 肖宇[1] 

机构地区：[1]湖南省交通科学研究院,湖南长沙410015

出　　处：《公路工程》2011年第5期44-46,51,共4页Highway Engineering

基　　金：湖南省交通科技项目[200718]

摘　　要：在车牌字符分割中由于泥尘、边框、铆钉、间隔符和车牌倾斜等因素影响,字符分割往往不准确甚至错误,针对这些问题提出了一种基于投影特征和先验知识的车牌字符分割算法。该方法先利用水平投影进行水平初分割,再利用垂直投影特征结合先验知识进行垂直分割,最后采用局部投影法实现水平精分割。实验表明该方法可较好的解决以上问题,分割准确率高且速度快,具有较强的实用性。Influenced by factors like mud, frame, rivet, space mark and plate slant, plate character segmentation tends to be inaccurate and even results in mistakes. In order to tackle these problems, this paper puts forward a plate character segmentation algorithm based on projection feature and prior knowledge. That is, firstly of all, to carry out horizontal coarse segmentation by horizontal projection, then use both vertical projection and prior knowledge to conduct vertical segmentation, at last, to carry out horizontal fine segmentation by local projection. The experiment demonstrates that such algorithm can solve the above mentioned problems accurately and practically with high-efficiency.

关 键 词：字符分割 投影特征 先验知识 

分 类 号：U491.1[交通运输工程—交通运输规划与管理]