基于简约码特性树的字母和数字识别

Recognition of Capital Letters and Arabic Numerals Based on Characteristic Tree of Reduce Code

针对由大写英文字母和阿拉伯数字组成的符号序列图像进行识别.字母和数字经细化处理,得到其骨架线,采用八方向链码及特定的编码跟踪规则形成由多条子链构成的唯一链码描述,提出了包含较少冗余信息、便于提取系列骨架线特征的简约码及简约码特征提取方法,并建立了对特征向量各分量酌情分开使用、形成分类特征树的策略,成功地解决了25个字母和10个数字的识别问题.测试结果表明,该方法及算法的平均正确识别率为98%,优于商用软件的识别率.

Image recognition of the character sequence, which is made up of capital letters and Arabic numerals, was investigated. Their skeletal lines were gotten by thinning disposal to the letters and numerals. The unique chain code description, which was made of some sub chains, was formed by Freeman chain code and specific coding rules. The reduce code and reduce code extraction, which is made up of little redundant information and is easy to extract a series of skeleton lines, were put forward. The strategy that distinguish the components of eigenvector and build the classifying characteristic tree was proposed. And the recognition problem 25 letters and 10 numerals was resolved. The testing result shows that the correct recognition rate, which is better than commercial software, is 98% in the method.