接触网支柱号牌定位与字符分割方法  

Location and character segmentation of catenary pole number plate

作　　者：闵锋[1,2] 吴涛 陈智羽[1,2] MIN Feng;WU Tao;CHEN Zhi-yu(School of Computer Science and Engineering,Wuhan Institute of Technology,Wuhan 430205,China;Hubei Province Key Laboratory of Intelligent Robot,Wuhan Institute of Technology,Wuhan 430205,China)

机构地区：[1]武汉工程大学计算机科学与工程学院,湖北武汉430205 [2]武汉工程大学智能机器人湖北省重点实验室,湖北武汉430205

出　　处：《计算机工程与设计》2020年第3期789-794,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61502354)。

摘　　要：针对高速巡检车成像质量较差的情况,提出一种基于形态学变换的定位算法,在此基础上,采用投影分割法提取字符。使用形态学变换得到号牌候选区域,利用其结构特征与HOG&SVM(梯度直方图与支持向量机)进行筛选,通过仿射变换与两次垂直投影提取号牌上下边界,利用投影矩阵的极值点得到候选分割点,分析其距离的变异系数得到合理的字符分割点。在普速铁路上的实验结果表明,号牌定位与字符分割正确率分别达到96.9%与95%,相对于传统垂直投影分割法与自适应投影分割法,字符分割正确率分别提升了19.9%与6.5%。In view of poor imaging quality of high-speed catenary inspection vehicle,a number plate location algorithm based on morphology transform was proposed.Based on this,a projection segmentation method was adopted to extract the characters of number plate.Morphology transform was performed to obtain some number plate candidate regions,the candidate regions were screened using the structure features and HOG&SVM(histogram of oriented gradients&support vector machine)method.The upper and lower boundaries of the number plate were extracted by affine transformation and two vertical projections.The candidate segmentation points were obtained by extreme points of the projection matrix,and the variation coefficient of the distances was analyzed to acquire reasonable character segmentation points.Experimental results on normal-speed railway show that the accuracy of pole number plate location and character segmentation are 96.9%and 95%respectively.Compared with the traditional vertical projection segmentation method and the adaptive projection segmentation method,the accuracy of character segmentation is increased by 19.9%and 6.5%respectively.

关 键 词：号牌定位 字符分割 形态学变换 投影矩阵 变异系数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]