基于字符边缘点提取的车牌定位方法  

License plate location based on character edge point extraction

作　　者：钟伟钊 杜志发 徐小红[2] 黄晓然[1] 朱同林[1] 

机构地区：[1]华南农业大学数学与信息学院农业多媒体技研究所,广东广州510642 [2]华南农业大学数学与信息学院,广东广州510642

出　　处：《计算机工程与设计》2017年第3期795-800,813,共7页Computer Engineering and Design

基　　金：高等学校博士学科点专项科研基金联合资助课题基金项目(20124404110018)

摘　　要：为快速精确地对图像中的车牌进行定位,提出一种基于车牌字符边缘点提取的车牌定位算法。使用Sobel算子对图像进行边缘检测,采用一种基于车牌特征的快速颜色特征提取算法;为排除其它干扰的边缘点,设计基于字符周围像素点比例的边缘点窗口扫描算法;使用高级形态学变换进行车牌候选区域提取,结合结构特征,使用基于像素点比例与水平跳变次数的排干扰法。对1000张不同环境下、不同颜色的车辆图像进行实验,实验结果表明,该方法成功率超过98%,在车身与车牌底色相同的情况下,可成功定位车牌。To locate the position of the license plate in the image fast and accurately,a license plate location method based on character edge point extraction was put out.An edge detection was carried out on the image based on Sobel and a rapid color feature extraction algorithm in view of the feature of license plate was put forward.To exclude the interference of other edge points,a window scanning algorithm based on the percentage of the surrounding pixels of the edge points was designed.The plate candidate region was extracted with advanced morphological transformation,the interference region was eliminated using the percentage of the pixels and change times in the horizontal direction combining with the structure feature.Tests on 1000 images in different environments with different colors were carried out.The results show that the success rates reach 98% above and the method locates the plate successfully when the color of the car body and the plate are same.

关 键 词：车牌识别 定位 颜色特征 边缘检测 结构特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]