摘要
针对当前车牌识别系统中图像分割方法应用单一,适用范围窄的问题,提出了综合应用多种图像分割方法的算法。车牌识别系统分为车牌图像预处理、车牌定位、字符分割、字符识别四个步骤。首先运用投影算法,把输入的原始车牌彩色图像转换成灰度图和二值图,并实现车牌位置的初步定位;其次运用OTSU算法,实现车牌的精确定位,并获取车牌的灰度图数据;然后运用动态自适应算法对车牌灰度图进行二值化;最后根据车牌的几何特征,对车牌字符进行分割及归一化处理。通过对4000张不同环境下车牌图片的测试,表明本处的算法可以成功的实现车牌定位及字符分割,具有较强的适应性,对后续的字符识别起到了重要作用。
In order to solve the problem of single image segmentation means in the system of license platerecognition,which is adapted poorly to license plate images with big dynamic range,the algorithm ofapplicating various image segmentation methods was put forward.The license plate recognition involvedfour components:the license plate image preprocessing,license plate location,character segmentation,and character recognition.At first,by use of the projection algorithm,the car license original color imageentered was converted to grayscale and binary map,candidate regions containing the license plateposition were got.Secondly,the OTSU algorithm was applied to get the precise position and gray data ofthe license plate.Then,dynamic adaptive algorithm was adopted for the binarization of the license plategray map.Finally,the characters were segmented and normalized by means of the geometric informationof the license plate area.The4000license plate images were tested,and the result shows that thearithmetic research is successful in the aspects of plate detection and character segmentation,which willplay an important role in the next character recognition.
作者
刘丽丽
Liu Lili(Department of Computer,Changzhi Institute,Changzhi Shanxi 046011,China)
出处
《科技通报》
北大核心
2017年第4期125-129,共5页
Bulletin of Science and Technology
关键词
投影
OTSU
动态自适应
车牌特征
车牌定位
字符分割
projection
OTSU
dynamic adaption
license plate features
license plate location
license plate recognition
