Computer Engineering and Applications ›› 2011, Vol. 47 ›› Issue (18): 204-206.

• 图形、图像、模式识别 • Previous Articles     Next Articles

Gradient features and region merging based vehicle license plate detection

PAN Wei,LIU Hongyu,AN Rong,YANG Nafei,HUANG Yijia   

  1. Information Engineering Institute,Capital Normal University,Beijing 100048,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-06-21 Published:2011-06-21

一种梯度特征与区域合并的车牌定位方法

潘 巍,刘宏宇,安 荣,杨娜菲,黄亦佳   

  1. 首都师范大学 信息工程学院,北京 100048

Abstract: The vehicle license plate detection is the important part of license plate recognition system.The license plate region is in the characteristics of gradient and frequent changes.High-pass filter is proposed to retain the gradient and frequent changes regions,merged adjacent regions as candidate regions after morphological processing.Combined with the geometric features of license plates and the ratio of regional objectives in the background to find the license plate location,the projection methods are adopted to remove the license plate frame,to achieve the exact position of license plate.Experimental results show that this approach can weaken the traditional requirements of the size of vehicle license plate,the image environment,shooting angles,etc.,and further improve the robustness and practicality.

Key words: license plate location, high-frequency characteristics, gradient, region merging

摘要: 车牌定位是车牌识别系统的重要组成部分。针对车牌目标所在区域梯度变换频繁的特点,利用高通滤波保留梯度变换频繁区域,形态学处理后,将相邻区域进行合并以确定车牌的候选区域。再结合车牌的几何特征与区域目标背景比,找到车牌位置,利用投影方法去除车牌边框,实现车牌的精确定位。实验结果表明,该方法削弱了传统车牌定位算法对车辆大小、图像环境、拍摄角度等的要求,进一步提高了算法的鲁棒性和实用性。

关键词: 车牌定位, 高频特征, 梯度, 区域合并