基于方向比重特征的维吾尔文笔迹鉴别方法  

Uighur Handwriting Identification Based on Directional Proportion Features

作　　者：谢鹏飞 卡米力.木依丁 

机构地区：[1]新疆大学信息科学与工程学院,乌鲁木齐830046

出　　处：《激光杂志》2014年第6期18-22,共5页Laser Journal

基　　金：国家自然科学基金资助项目(61065001);新疆少数民族科技人才特殊培养计划项目(201023116)

摘　　要：笔迹鉴别是一种有效的身份鉴别方法。作者针对维吾尔文文字书写特点提出了一种文本无关的笔迹鉴别方法,该方法采用递推方式,首先,以边缘图像为基础,利用8邻域像素点,计算边缘像素点四维方上的方向比重;然后,获取微切分窗口内所有的边缘像素点的方向比重;最后,求得窗口中书写笔迹的方向比重,并以此作为特征结构进行笔迹鉴别。实验不但改进了笔迹图像预处理,而且根据特征结构的特点,在加权卡方距离度量度量方法上进行笔迹密度再加权计算。结果表明,该方法提高了鉴别效果,有很强的实用性。Handwriting identification is an effective method of identification.According to the features of Uygur letters, this paper proposed a text-independent handwriting identification technique. Based on edge image, the author used a recursive system. Firstly, we counted the edge pixel＇s 8-neighborhood pixels to get its proportion of four dimensional directions. Secondly, did this process on all the edge pixels of local windows.At last, obtained the handwriting＇s proportion of directions which we made as feature structures.The method made improvement in image＇s pre-preoccessing. According to the the characteristics of feature structures,the author also used the density of stroke to weight weighted Chi-square distance formula again to measure the similarity between handwritings. With a strong practicability, the method achieve a better identification result.

关 键 词：维吾尔文 笔迹鉴别 文本无关 8邻域像素点 笔迹密度 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]