摘要
为了提高车牌上的字符识别准确率,提出一种结合支撑矢量机(SVM)和小波的字符识别方法.通过对字符图像水平和垂直两个方向的投影曲线分别进行小波分解,得到投影曲线的近似表示.在近似曲线中提取字符的特征参数,用这些特征参数构成特征矢量作为SVM训练和分类的基本参数,再将特征矢量输入支撑矢量机网络训练,最后通过树型分类识别模型识别字符.实验仿真表明,该字符识别方法的平均准确率为97.15%,平均识别速度为每个字符19.15 ms.
To improve the recognition accuracy of the characters on car license plate, a novel method based on support vector machine （SVM） and wavelet decomposition was proposed. The projective curves along the horizontal and vertical directions were obtained from the character images, and were approximated by using wavelet decomposition. The characteristics of the character images were then extracted from the approximated curves, and were taken as the fundamental parameters for SVM training and classification. Characters were classified and recognized by the use of decision-tree classification-recognition model with high recogniton capability. Experimental results show that the average accuracy rate is about 97.15% and the average recognition speed is about 19.15 ms with the proposed method.
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2005年第12期2016-2020,共5页
Journal of Zhejiang University：Engineering Science
