基于HSV空间颜色和纹理特征的车牌定位算法研究  

Research on License Plate Location Algorithm Based on HSV Spatial Color and Texture Features

作　　者：赵洋 闵升锋 李大舟 ZHAO Yang;MIN Sheng-feng;LI Da-zhou(Shenyang University of Chemical Technology,Shenyang 110142,China)

机构地区：[1]沈阳化工大学计算机科学与技术学院,辽宁沈阳110142

出　　处：《沈阳化工大学学报》2022年第1期90-95,共6页Journal of Shenyang University of Chemical Technology

基　　金：辽宁省博士启动基金项目(201601196)。

摘　　要：车牌识别系统中的一个关键性技术是车牌定位,其准确性直接影响后面的车牌字符分割和识别的效果.当背景与车牌区域颜色相似时,在颜色车牌定位算法中不能正确定位出车牌.针对上述问题,提出一种基于HSV空间颜色和纹理特征的车牌定位算法.该算法首先通过HSV空间变换,使颜色空间具有独立性,从而分割出不同颜色区域;其次,通过车牌区域字符集中且边缘信息丰富的特点,设置字符与字符边缘之间的梯度跳变次数的阈值,在复杂的背景下判断待测试区域是否为车牌区域;最后,通过二值化直方图投影法定位并分割出车牌区域,建立BP神经网络训练识别车牌字符来验证算法的有效性.实验结果表明:该方法能够在颜色相似和背景复杂的情况下正确定位车牌区域,在相同的条件下对比其他定位算法,其识别率达95.06%.A key technology in license plate recognition system is license plate location, and its accuracy directly affects the segmentation and recognition of the following license plate characters.When the background is similar to the color of the license plate area, the license plate location algorithm cannot be located correctly in the color license plate location algorithm.To solve the above problems a license plate localization algorithm based on HSV spatial color and texture features is proposed.Firstly, the algorithm makes the color space independent through HSV space transformation, so as to segment different color regions.Secondly, according to the character set and rich edge information of the license plate area, the threshold of the gradient jump times between the character and the character edge is set to judge whether the area to be tested is the license plate area under the complex background.Finally, the license plate area is located and segmented by the binary histogram projection method, the BP neural network is established, and the license plate characters are trained to verify the effectiveness of the algorithm.The experimental results show that this method can correctly locate the license plate area under the condition of similar color and complex background.Compared with other location algorithms under the same conditions, the recognition rate is as high as 95.06%.

关 键 词：车牌定位 HSV空间 纹理特征 梯度跳变 二值化直方图 BP神经网络 

分 类 号：TP391.413[自动化与计算机技术—计算机应用技术]