采用特征点提取算法的车牌倾斜校正方法研究  

A Method of Slant License Plate Correction Based on Feature Point Detectors

作　　者：李莹[1] 秦丽娟[1] 王慧科 刘小芳[1] 李东昇[1] 

机构地区：[1]沈阳理工大学信息科学与工程学院,辽宁沈阳110159

出　　处：《沈阳理工大学学报》2014年第6期1-6,共6页Journal of Shenyang Ligong University

基　　金：国家自然科学基金资助(61203163);国家重点实验室基金资助(2013-006)

摘　　要：针对车牌倾斜校正问题,首先简介了四种特征点提取的算法,并结合实验结果分析了各算法的优缺点,进而提出了比传统算法更有效的基于非线性双边滤波器的Harris算法,然后提取出车牌区域的字符角点信息,再结合惯性主轴算法对倾斜车牌进行校正。实验结果证明,改进后算法有效地抑制了伪角点数目,提高了车牌角点检测的精度,使得车牌倾斜校正的精确度和实时性大幅度提高。For the problem of slant license plate correction,firstly,the four algorithms of feature points extracted are introduced and the advantages and disadvantages of the algorithms are analyzed by experimental results. And then,traditional algorithm of Harris corner detection is improved and the corner point information of character from the license plate region is extracted with corner detection based on a nonlinear bilateral filtering. Slant license plate is corrected by Harris corner detection improved and the inertial principal axis algorithm. Experimental results showthat false corner number is effectively suppressed by the improved algorithm. As a result,the precision of license plate corner detection,the accuracy and real time of the slant plate correction are improved.

关 键 词：倾斜校正 特征点提取 双边滤波 HARRIS角点检测 惯性主轴 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]