基于图像特征的汉字字形相似度计算  

The Similarity Calculation of Chinese Character Glyph Based on Scale-Invariant Feature Transform

作　　者：王昭 杨婧 杨敏 Wang Zhao;Yang Jing;Yang Min(Shanxi Branch of National Computer Network Emergency Response Technical Team/Coordination Center of China,Taiyuan Shanxi 030012,China)

机构地区：[1]国家计算机网络应急技术处理协调中心山西分中心,山西太原030012

出　　处：《山西电子技术》2024年第2期61-63,共3页Shanxi Electronic Technology

摘　　要：汉字作为一种数量众多,结构复杂的象形文字,其字形的相似度计算对汉语的错别字识别和纠正具有重要意义。将汉字视为图形,通过尺度不变特征变换算法提取汉字的特征点,并利用汉字间特征点的对应关系,提出了一种基于特征点的相似度计算方法。对《通用规范汉字表》的8105个汉字进行了相似度分析,结果表明,该方法和人的认知结果比较一致。As a pictograph with large number and complex structure,the similarity calculation of Chinese characters is of great significance for the recognition and correction of misspellings in Chinese.In this paper,Chinese characters are treated as graphs,feature points of Chinese characters are extracted by Scale Invariant-Feature Transform(SIFT)algorithm,and a similarity calculation method based on feature points is proposed by using the corresponding relationship between feature points between characters.The similarity analysis is carried out for 8105 Chinese characters and the results show that this method is more consistent with the human cognitive results.

关 键 词：汉字字形 形似字 相似度计算 尺度不变特征变换 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]