摘要
为了提高复杂背景下车牌定位的准确性,在分析各类车牌定位算法的基础上,提出了一种结合纹理特征和视觉词包模型的多信息定位方法.对二值化后的图像中的纵向纹理横向膨胀,生成车牌候选区域,根据车牌的长宽比对车牌进行粗定位.利用视觉词包表示粗定位后的矩形轮廓,使用支持向量机分类确认车牌矩形区域,精确定位出车牌位置.该方法对140张测试样本定位的准确率为96.4%,抗干扰性强.
In order to increase the accuracy of license plate location in a complex background,the paper presents a multi information locating method with morphological processing and bag of visual words model based on the analysis of all kinds of a license plate location algorithms.Then the rectangular regions that match the aspect ratio of a license plate are represented by a histogram over visual words based on the bag of visual words.The method can locate the license plate clearly and accurately in the rectangular region by the classifying function that Support Vector Machine resolves.The method has high identification accuracy strong anti-interference performance 140 samples were tested by this method,and the accuracy rate is 96.4%.
出处
《西安工业大学学报》
CAS
2016年第5期382-387,共6页
Journal of Xi’an Technological University
关键词
车牌定位
复杂背景
视觉词包
支持向量机
车牌纵向纹理
multiple license plate position
complex background
bag of visual words
license plate vertical texture