节点文献

印刷体文字识别的研究

Research on Printed Chinese Character Recognition

【作者】 倪桂博

【导师】 张国立;

【作者基本信息】 华北电力大学(河北) , 计算机应用技术, 2008, 硕士

【摘要】 笔划代表着汉字的内部特征,笔划穿越次数是对笔划进行全穿越,反应了汉字的整体特征,全穿越在粗分时区分汉字的能力不是太强,增加了二级识别的工作量。本文除了提取笔划全穿越外还提取笔划半穿越,并把半穿越的次数进行重新组合形成新的特征值。把全穿越和半穿越结合起来作为汉字的特征值,对汉字进行粗分,粗分不能区分的汉字,采用四个角的能量值密度特征对汉字进行细分。实验结果表明了该方法的有效性。与单独使用全穿透方法相比,本文提出的方法在粗分时区分汉字的能力增强,减少了二级识别的工作量。本文还对印刷体数字进行了研究,从数字的结构形状着手,通过分析印刷体数字的形状,提出了一种基于结构形状的印刷体数字识别方法。该方法不用对字符图像进行复杂的细化处理,减少了因细化带来的误差问题,因而识别速度非常快,实验证明了该方法的有效性。

【Abstract】 Stokes represent internal character of Chinese Character, The previous method of traversing times of strokes is full-breakthrough to stroke, but this method is not effective for some Chinese Characters. This paper introduces half-breakthrough of strokes, and makes traversing times combine newly then obtains a new feature. It is used to be the first recognition with the combination of full-breakthrough and half-breakthrough. If it can not be recognized then make the second recognition with energy-density. This method does not need to complex thin to the character picture, reducing the erroneous question which is brought by thinning。The result shows this method is effective. The effect of the new method has obvious progress compared with the full-breakthrough only in first recognition, decreasing workload of the second recognition. A method of printing digital has been proposed that based on the structure shape, through analyzing the structure shape of the printing digital. This method does not need to complex thin to the character picture, reducing the erroneous question which is brought by thinning, so the recognition speed is quickly. The result shows this method is effective.

【关键词】 笔划穿越次数能量值汉字识别
【Key words】 stroketraversing timesenergyChinese Character Recognition
  • 【分类号】TP391.43
  • 【被引频次】12
  • 【下载频次】607
节点文献中: 

本文链接的文献网络图示:

本文的引文网络