| 识别手写印刷体汉字的二维扩展属性文法方法 |
| 赵明
|
| 1988
|
| 学位授予单位 | 中国科学院软件研究所
|
| 学位 | 博士
|
| 学位授予地点 | 中国科学院软件研究所
|
| 摘要 | 本论文研究手写印刷体汉字识别方法，主要的工作为：1. 提出了二维扩展属性文法模式识别方法，可实现自顶向下推导和自底向上归约双向信息传递和控制。2. 提出了一种直接利用二维信息进行识别的方式，可避免特征线性化造成的信息丢失，易于用并行算法实现。3. 提出了多义文法、共生文法和结构推断三种在文法之间建立联系，利用结构类比区分极相似字的识别算法。4. 提出了多冗余归约机制，双边缘弹性跟踪笔段抽取算法，多冗余有引导部件抽取算法，等。5. 用2-D EAG方法对100字种实际手写汉进行了识别实验，在识别大畸变汉字（包括部分连笔字）方面得到了很好的结果。 |
| 其他摘要 | This paper is on recognition method of hand printed Chinese characters. The main aspects of the paper are: 1. Raise the pattern recognition method of Two-Dimensional Extended Attribute Grammars(2-D EAG for short). In this method, both top-down deduction and bottom-up reduction are realized for two-way information transmitting and control. 2. Propose a recognition scheme utilizing directly two dimensional information, by which the information loss caused by linearization of features can be avoided and parallel algorithm can be easily adopted. 3. Provide polysemous grammars, coexisting grammars and structure inferrings which build special linkages among grammars and dikstinguish among similar samples by structural analogy. 4. Besides, realize also redundant reduction scheme, stroke segment drawing algorithm with elastic tracing of contour lines, redundant guided component drawing algorithm, etc. 5. recognition experiment with 2-D EAG method for 100 categories of hand printed Chinese characters is conducted and very good result is achieved in overcoming large scale structural distortion (including some cursive writings). |
| 页数 | 51
|
| 语种 | 中文
|
| 内容类型 | 学位论文
|
| URI标识 | http://ir.iscas.ac.cn/handle/311060/7098
|
| 专题 | 早期
|
推荐引用方式
GB/T 7714 |
赵明. 识别手写印刷体汉字的二维扩展属性文法方法[D]. 中国科学院软件研究所. 中国科学院软件研究所,1988.
|
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。
修改评论