一种新的货运列车车号分割算法  

A New License Plate Segmentation Algorithm of Freight Train

作　　者：牛智慧[1] 赵歆波[1] 葛莉[1] 

机构地区：[1]西北工业大学计算机学院,陕西西安710129

出　　处：《计算机技术与发展》2014年第8期18-21,共4页Computer Technology and Development

基　　金：国家自然科学基金资助项目(61171156);西北工业大学研究生创业种子基金(Z2013138)

摘　　要：在对现有的货运列车车号分割算法及相关字符分割算法对比研究的基础上,文中提出并实现了一种新的货运列车车号分割算法。根据上下轮廓特征初步确定车号字符串图像的候选分割位置,然后根据字符尺寸比例和数字的弧特征,对断裂字符进行合并和对粘连字符进行再分割。该方法巧妙地避免了传统的投影分析分割法中处理粘连字符的难题,也避免了噪声对连通域的影响。与传统方法相比,具有较好的鲁棒性,达到了较高的精度和运行效率,为整个车号识别系统的精确性和稳定性提供了保障。Through the comparative research of the existing segmentation algorithms of license plate and the related segmentation algorithms,a new segmentation algorithm of license plate of the freight train is proposed and implemented. According to the upper and lower contours of the license plate numbers,the algorithm yields a list of candidate segmentation locations,then merges the break characters and resegments the touching characters depending on the character size ratio and arc features of numbers. This method neatly avoids the nightmare of segmentation of the touching characters in the traditional projection analysis,and also avoids the impact of noise on the connected fields. Compared with traditional methods,this method has better robustness,achieving a higher accuracy and efficiency and providing a guarantee of accuracy and stability of license plate recognition system.

关 键 词：车号分割 轮廓特征 弧特征 粘连字符 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]