车牌字符特征提取与识别应用

Feature Extraction and Recognition Application of License Plate Character

基于仿生模式识别思想,对车牌字符这一有限集,以小波分析高频子图为载体,摆脱传统特征基于字符二维平面的统计、结构特征,提出高维超平面特征概念,设计高维超平面滤波器,感兴趣于字符部分最为重要的特征,并对其进行认知编码,很大程度上排除了传统识别时整体提取特征的信息的混淆和冗余,降低了误识、拒识率,达到很好的实用指标。

Based on biomimetic pattern recognition thoughts, as the carrier with wavelet analysis high frequency subgraphs, the concept of high-dimensional hyper-plane features is proposed for the limited set of characters on license plate. High-dimensional hyper-plane filter is designed. The most important features of character portion are cognized and coded. Confusion and redundant information of the traditional overall recognition feature are largely excluded. The false consciousness and rejection rates are reduced. Good practical indicators are achieved.