基于HSI空间的颜色算法提牌照识别的性能  

Improving license plate recognition system by color algorithms based on HIS space

作　　者：傅一平[1] 李志能[1] 袁丁[1] 

机构地区：[1]浙江大学信电系,浙江杭州310027

出　　处：《计算机工程与设计》2004年第5期703-707,共5页Computer Engineering and Design

摘　　要：为了研制高性能的车辆牌照识别系统,在原有基于灰度图牌照分析的基础上,考虑利用牌照的颜色信息来提高牌照识别的性能,提出了HSI空间上确定彩色牌照颜色的方法,进而用改进的基于K-means的HIS空间颜色聚类算法得到背景色范围,然后用数字形态学方法对聚类后的牌照背景进行一系列的形态运算,去除聚类后得到的前景字符对背景的干扰,最后,用边界颜色拟和直线确定精确的牌照边界,并实现对倾斜牌照的矫正。对大量的彩色车牌图片进行实验结果表明,利用HSI空间的牌照颜色信息可以提高整体牌照识别系统的性能。In order to improve the performance of license plate recognition system,the color of license plate was considered fol- lowing the analysis of the gray.First,a new method for determining the color of license plate was presented based on HIS color space.And then an improved k-means color clustering algorithm on HIS space was provided to position the background of license plate.And then by further morphological operations,the forground characters which blured the background was removed.Finally, boundary lines of license plate based on the background were determined,which rectified the lean license plate.Experiment result shows that the color algorithms of license plate based on HIS space can improve the whole performance of license plate recognition system greatly.

关 键 词：HSI空间 牌照识别 颜色聚类算法 数字形态学 光学字符识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]