摘要
提出了一种基于混合轮廓特征的压印字符识别新方法。首先提取字符轮廓的矩不变量和Fourier描述子,然后将2种特征融合构成混合特征,最后将新特征输入BP网络进行识别。实验证明,混合轮廓特征不仅克服了2种特征单独使用时的缺点,提高了识别率,而且可以提高特征的平移、缩放以及旋转不变性。
We proposed a novel recognition method for the pressed characters based on the mixed outline features. First, the moment invariants and the Fourier descriptors are extracted, then, the two kinds of features are fused to form mixed features;finally, the mixed features are put into the BP network to implement the recognition. The experiments proved that the mixed features can overcome the disadvantages of the two features used separately, improve the recognition rate, and enhance the features＇ invariability aganinst image rotation,scale change and translation.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2007年第10期1244-1247,1259,共5页
Journal of Optoelectronics·Laser
基金
教育部博士点基金资助项目(20060422011)
