手写印刷体汉字的笔段抽取及偏旁识别

本文采用对汉字点阵图象进行方向变换的方法抽取汉字的笔段,采用结构分析的方法识别分布于汉字四周的偏旁,对国标一级汉字申的90类偏旁计一万余字进行了偏旁抽取试验,当侯选偏旁数<5时,累计正确侯补率>96％。

In the paper, a method for stroke extraction of Chinese characters is presented by means of direction transformation of dot matrix of Chinese characters. With structure analysis, the peripheral ridcals are effectively recognized. For 99 different radicals in national standard level 1 Chinese characters, an experiment shows that the correct recognition rate is greater than 96% when candidate number amounts to three.