基于多特征提取的手写汉字识别算法及其实现

基于多特征提取的手写汉字识别算法及其实现
目   录中文摘要 Ⅰ
ABSTRACT Ⅱ
目录 III
图片索引目录 Ⅳ
绪论 1
1 汉字识别原理及常用算法 4
  1.1 汉字识别的原理 4
  1.2 基于统计决策方法的汉字识别 5
  1.3基于结构特征方法的汉字识别 6
2 基于多特征提取的识别算法 9
  2.1手写汉字识别的现行算法分析 9
 2.1.1以笔划序列为特征的识别算法 9
 2.1.2基于统计特征的汉字识别算法 11
 2.2现行算法的结合和改进 12
3 识别算法对应的VC代码 14
  3.1输入预处理 15
  3.2识别算法 18
4 算法实现结果和分析 20
  4.1 模拟结果介绍 20
  4.2 结果分析 21
5 总结与展望 24
  5.1 论文总结 24
  5.2 算法的前景展望 24
参考文献 25附:图片索引目录图1.1 汉字识别原理框图 4
图 1.2 汉字图像的粗外围图 5
图 2.1 二维平面的划分图 9
图 2.4 汉字图像的统计特征图 11
图 2.6 汉字识别算法流程图 13
图 4.1 程序运行界面 20
图 4.2 汉字的写入 20
图 4.3 系统识别写入的汉字图像 21
图 4.4 通过“学习”过程向字库中添加汉字特征 21
图 4.5 能够识别的汉字图像示例 22
图 4.6 不能识别的汉字图像示例 22
ABSTRACT  The article proposes an algorithm for hand-written Chinese character recognition and its realization . The algorithm is based on the multi-characteristics, it unifies and improves the algorithm based on the statistic characteristic and the algorithm based on the structure characteristics.
  The algorithm adopts the classified distinction through the match distinction stage. It takes down the character image’s input unique feature when the image input. With the input unique feature, it can obtains the pen delimits characteristics of the unknown character image. The first classification can be managed based on these steps, thus the range of the second classification will smaller. After the first classification, it must withdraws the statistic characteristics of the image. The algorithm adopts the grid division method in this step. It makes a division to the image and then compute the element ratio of every areas. As a result, the character image can be recognized by the steps above.
  The recognition algorithm based on multi-characteristics is extremely effective to the Chinese character recognition.The recognition rate can achieves above 90% .This article also provide methods for the characters written standardless. Key words: Writing recognition, Characteristic withdraws, Pattern matching, Pattern recognition
摘 要
      该文提出了一种基于多特征提取的手写汉字识别算法及其实现。该算法同时运用了基于统计特征和基于结构特征的汉字识别算法,并对以上两种算法做了适当的改进。
算法在匹配判别阶段采用了分类判别思想,在汉字图像输入的同时,记录汉字图像的输入结构特点,在此基础上得到输入的未知汉字的笔划特征,由笔划特征即可对未知汉字做粗分类,这样可以缩小判别时的汉字范围。经过汉字粗分类后,提取整个汉字图像的统计特征,该步骤采用了网格划分方法,即先对图像做划分,再计算每一网格内文字相素与整个图像中文字相素的比值,通过这些特征对文字图像做细分类即可得到识别结果.
  基于多特征提取的识别算法对汉字的识别十分有效,能使识别率达到90%以上,该文同时也给出了对书写不规范的字的解决方案。
关键词:文字识别,特征提取,模式匹配,模式识别809

[1] [2] [3] [4] [5] [6] [7] [8] 下一页

Copyright © 2007-2012 www.chuibin.com 六维论文网 版权所有