摘要
票据打印字符识别有着广泛的应用前景。但由于票据打印字符本身所具有多噪声、字体及字符大小多变等特点,导致了票据打印字符识别始终是一项技术难题。提出了一种基于粗糙集理论的票据打印字符识别系统,首先采集打印数字字符构成训练样本集,提取训练样本的特征向量,建立样本集的特征矩阵。然后基于粗糙集理论,采用一种有效的属性集约简方法和一种新的规则提取方法,从中导出三个规则集合,基于该三个规则集对测试样本分别进行决策,最后对决策结果进行决策级融合。将该方案应用于纸制飞机票上打印数字字符的识别上,实验证明了该方法的可行性及有效性。
According to the inherent properties of characters recognition, we propose a comprehensive decision scheme with multiple templates on the basis of rough set theory in the paper. As to concrete realization, we adopt an effective method for removing redundant attributes and a new approach for extracting rules, which directly focus on a primitive training template, and then recognize the test samples utilizing three templates produced finally. It is showed by the experimental result, this decision scheme is very practical and effective
出处
《计算机与数字工程》
2008年第8期70-73,共4页
Computer & Digital Engineering
关键词
粗糙集
属性约简
规则提取
字符识别
rough set, extract attributes, extract rules, characters recognition
作者简介
郭森，男，讲师，研究方向：图像处理，模式识别。
