摘要
针对复杂场景下车牌识别问题,提出了一种新的车牌识别方法.该方法通过颜色空间划分预分割车牌目标,然后采用矩形特征匹配定位符合车牌形状的候选块,并对候选块进行几何校正和自适应分割,再将分割出来的每个字符进行归一化特征提取,最后通过字符的积分特征逐一对每个候选块进行识别,识别成功者即为真实的车牌.实验表明,该方法能适应大多数复杂的场景,不局限于图像的分辨率,以及拍摄的角度,保证了车牌识别的准确率.
A new plate recognition method is proposed for complicated background.It uses color space partition to segment the plate object,then adopts the rectangle matching algorithm to locate the multiple candidate plate blocks.And geometric correction as well as adaptive segment algorithms are applied to the candidate plate blocks,then the divisional single characters must be normalization and stable features are extracted.Last,it determines the true plate block and recognizes each character based on character integration features matching.Experiments indicate that proposed plate recognition method is suit for majority complicated background.Furthermore,this method is not restrict for image resolution together with capture angle,promotes the plate recognition＇s accurate.
出处
《湘潭大学自然科学学报》
CAS
北大核心
2016年第1期90-92,共3页
Natural Science Journal of Xiangtan University
基金
四川省高校创新团队项目(13TD0001)
关键词
图像处理
颜色空间
车牌识别
阈值分割
积分特征
image process
color space
plate recognition
threshold segment
integration feature
