摘要
提出了结合空域拓扑结构和频域特征的车牌定位实用算法。在空域上,首先对原始照片进行预处理,包括灰度拉伸处理从而均衡光照影响,有利于增强算法的鲁棒性;然后对前面处理结果进行形态学梯度变换从而加强车牌文字区域的线状结构,有利于突出文字的频率特征。在频域上,利用车牌文字的横向频率特征,按照一定稠密度条件筛选水平象素,得到若干相对稠密的水平点集。接着在空域上对频域处理结果进行几何归并、拓扑筛选。最后利用车牌本身特点提出以一种快速倾斜纠正方法,并经过投影过滤,从而实现车牌区域快速准确定位。
A utilitarian locating plate method is presented, in which frequency and space topological characteristics about Plate area are considered, Firstly, origin image was processed by gray expanding so that the effects of lighting was averaged and the algorithm was more robust; and the result was handled by morphology gradient operator so that linear shapes of plate area and frequency characteristics of plate were highlighted. Then frequency operator was defined according to havezening frequency characteristics of plate, and some valuable horizontal sets were produced. Thirdly, the horizontal sets were combined geometrically and filtered by geometrical topological. Finally, the slope candidate rectangles patches were corrected and projected vertically. As a result, the correct position of plate comes out efficiently.
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2005年第10期2349-2351,2357,共4页
Journal of System Simulation
关键词
车牌定位
拓扑结构
灰度拉伸
形态学
垂直投影
License Plate
topological structure
gray expanding
morphology
vertical projection
