基于CPD算法与笔段权重的楷体字笔画提取研究  

RESEARCH ON CHINESE CHARACTER STROKE EXTRACTION BASED ON CPD ALGORITHM AND SEGMENT WEIGHT

作　　者：刘相聪 李壮峰 姜杰[1] 李艺 Liu Xiangcong;Li Zhuangfeng;Jiang Jie;Li Yi(College of Education Science,Nanjing Normal University,Nanjing 210023,Jiangsu,China)

机构地区：[1]南京师范大学教育科学学院,江苏南京210023

出　　处：《计算机应用与软件》2022年第8期204-212,219,共10页Computer Applications and Software

基　　金：国家社会科学基金“十三五”规划2016年度教育学课题(BCA160052)。

摘　　要：为了解决传统汉字笔画提取效率较低,使用点集匹配进行笔画提取成功率不高且难以纠错等问题,提出一种基于点集匹配与笔段权重的楷体字笔画提取算法。利用一致性点漂移算法(Coherent Point Drift,CPD)对于字集进行映射,根据映射后的结果分配笔画信息;利用关键点进行笔段分割,再使用第一次笔画分配后笔段中包含的笔画信息修正笔画匹配结果,最终获得更为贴切的汉字笔画分配结果。实验结果表明,改进后算法对于楷体字的笔画提取成功率达到97.5%;针对三种不同楷体字的壁画提取成功率为89.17%。与传统笔画提取方法相比,该算法的汉字笔画提取成功率有所提升而且针对不同的提取字体的笔画提取普适性得到增强。In order to solve the problems of low efficiency of traditional Chinese stroke extraction,low success rate of stroke extraction using point set matching and difficulty in error correction,a Chinese stroke extraction algorithm based on point set matching and stroke segment weight is proposed.We used coherent point drift(CPD)to map the character set,assigned stroke information according to the result of mapping.We used key points to segment the stroke,then used the stroke information contained in the first stroke allocation to correct the stroke matching result,and finally obtained more appropriate stroke matching results.The experimental result shows that the improved algorithm achieves 97.5%of the stroke extraction success rate for regular script,and 89.17%of the mural extraction success rate for three different types of regular script.Compared with the traditional stroke extraction method,the improved algorithm improves the success rate of Chinese character stroke extraction and enhances the universality of stroke extraction for different extracted fonts.

关 键 词：笔画提取 点集匹配 笔段权重 汉字骨架 笔段划分 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]