摘要
针对传统的基于骨架特征识别方法具有运算量大、速度慢、对细化等预处理要求高,以及受起始点、断笔、跟踪方式影响极大等缺点,本文提出一种基于凹凸特性笔顺编码的识别方法。该方法首先计算数字图像的赋值背景场,再从中提取凹凸特性,然后根据凹凸特性进行笔顺编码,最后将编码与正则表达式表匹配。该方法不需要细化,减少了细化形变可能带来的误识和拒识;也不需要搜索、拟合等复杂处理,因此简单快速,运算量小。实验表明,该方法能大大提高识别率和速度。
Traditional methods for the off-line recognition of hand-written digits based on skeleton coding have such defects as low speed, huge computation,and so on. The paper proposes a new method based on convex-concave-coding, which only finds the circle and convex areas from the value-associated background, and then combines and codes them. It needs no thinning, which reduces the possibility of errors and reiections because of the thinning distortion, and it needs no complex skeleton tracing and curve fitting, which means it is simple and fast. Experiments show that it can improve the recognition rate and speed rapidly.
出处
《计算机工程与科学》
CSCD
2007年第5期69-70,108,共3页
Computer Engineering & Science
关键词
凹凸特性
手写体数字
笔顺编码
convex and concave feature
handwritten digits
sequence coding
