Journal of Information Security Research ›› 2016, Vol. 2 ›› Issue (1): 58-65.

Previous Articles     Next Articles

A Colors Based Algorithm for License Plate Location

Wu Weixi   

  • Received:2015-12-18 Online:2016-01-05 Published:2016-01-18



  1. 北京信息科技大学信息管理学院信息安全系
  • 通讯作者: 吴惟希
  • 作者简介:吴惟希 硕士研究生,主要研究方向为机器学习与人工智能. 赵刚 博士,副教授,主要研究方向为人工智能与信息安全.稿件负责人:赵刚13718677515通信作者:赵刚(

Abstract: Nowadays processing methods for license plate usually convert photos to gray scale images at first, and then find the characteristics based on the character of the edge texture for positioning. These methods do not work well on some conditions that there are the disturbance near the license plate, the instability of the light illumination and character edge gradient caused by floating dust. Then if it is in the fog and haze, these methods are more powerless. For that, in this paper a color based location algorithm is proposed, which is to select the candidate region according to the pixel color similarity and color domain in different color spaces and then based on a variety of geometric features to select the license plate region. This algorithm realizes the location of being at different light conditions, the strong interference of complex background and damaged license plates. Even in the fog and haze, this method performs excellent for those extremely blurred image.

Key words: location, color space, color similarity, Manhattan distance with weight, illumination condition, interference, fog and haze

摘要: 现今的车牌图像处理方法一般是先把图像转为灰度图像,再基于字符的边缘纹理来找寻特征进行定位.这类方法对于图像内车牌附近的干扰不能较好处理,对不稳定的光照条件及浮灰导致的字符边缘梯度不大的情况处理效果也欠佳,在大雾天更是无能为力.对此,提出了一种基于色彩的定位算法,在不同色彩空间根据像素色彩相似度及基色域来选出候选区域,再依据多种几何特征筛选出车牌区域.算法实现了不同光照条件下及复杂背景强烈干扰下的定位,以及对受损车牌的定位,并在处理处于雾霾天气下极度模糊的图像也能表现优异.

关键词: 定位, 色彩空间, 色彩相似度, 带权值的曼哈顿距离, 光照条件, 干扰, 雾霾