基于色差的车牌快速定位算法研究  

RESEARCH ON FAST LOCATION ALGORITHM OF LICENSE PLATE BASED ON COLOR DIFFERENCE

作　　者：郑楷鹏 郑翠环 郭山红[1] 陈康宇 

机构地区：[1]南京理工大学电子工程与光电技术学院,江苏南京210094

出　　处：《计算机应用与软件》2017年第5期195-199,共5页Computer Applications and Software

摘　　要：针对现有车牌定位算法的抗干扰能力弱和速度较慢问题,提出基于色差的车牌快速定位算法。设计了提取蓝色、黄色和白色像素点的色差公式,将车辆RGB图像转换到色差空间。利用迭代法对色差图像进行阈值分割得到二值图像,并利用形态学处理和标记连通域,最后结合车牌长宽比特征去除干扰定位车牌。对实际车辆视频定位处理结果表明,提出的算法可以实现白天和夜晚蓝牌车和黄牌车的车牌准确定位,定位准确率分别达到95.1%和92.5%,定位平均耗时0.026 s,优于传统的基于HSV、HIS和YUV车牌定位算法。处理结果表明该定位算法可在实际普通道路交通中实现实时准确定位。Aiming at the problem of weak anti-interference ability and slow speed of the existing license plate locationalgorithm , a fast location algorithm of license plate based on color difference is proposed. The color difference formula of the blue , yellow and white pixels is designed to convert the RGB image into the color difference space. The iterative method is used to obtain the binary image by threshold segmentation, and the morphological processing and the mark connected domain are used. Finally , combined with the license plate aspect ratio feature to remove interference and locate the license plate. The results of real vehicle location show that the proposed algorithm can accurately locate the license plates of daytime and night blue and yellow cards with the accuracy of 95. 1 % and 92.5% respectively , and the average positioning time is 0. 026 s , which is superor to the traditional based on the HSV , HIS and YUV license plate location algorithm. The results show that the algorithm can achieve real-time and accurate location in practical road traffic.

关 键 词：车牌定位 色差 阈值分割 形态学处理 标记连通域 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]