一种基于SVG的在线连续手写汉字多步分割方法*

基于可伸缩矢量图SVG的在线手写汉字是以SVG图像作为汉字图像格式、以SVG的path对象作为笔画的基本存储单元来对汉字进行显示和存储的，笔画的轮廓是以手写过程中记录的坐标值作为特征数值加以确定的。基于此种SVG手写汉字存储和表示形式，本文提出一种基于图论的在线连续手写汉字多步分割方法。该方法根据汉字笔画间的坐标位置关系对手写笔画序列构建无向图模型，并利用图的广度优先搜索将原笔画序列分割为互不连通的笔画部件，使偏旁部首分离较远、非粘连汉字得到正确分割；然后利用改进的tarjan算法对部件中的粘连字符进行分割，最后基于笔画部件间距，利用二分类迭代算法对间距进行分类，找出全局最佳分割位置，对过分割的部件进行重组合并。实验结果表明，该方法对于在线手写汉字的分割是有效可行的。

A Multi-step Segmentation Method of Online Handwritten Chinese Characters Based on SVG

Online handwritten Chinese characters based on SVG is uses SVG image as Chinese character image format,and SVG path object as the basic storage unit stroke to display and storage Chinese characters.And the outline of the stroke is determined by the coordinate values recorded in the writing process .Based on this SVG handwritten Chinese character storage and representation method,we put forwards a method of online continuous handwritten Chinese character multi-step segmentation based on graph theory.Firstly,this method constructs an undirected graph model of the sequence of handwritten strokes based on the coordinate position relations among Chinese characters, uses the breadth-first search of the graph to divide the original stroke sequence into non-connected stroke parts,to let that radical separation of radical far, non-adhesive Chinese characters get right segmentation.and then uses the improved tarjan algorithm to segment this sticky characters in the parts.Finally,based on the spacing between stroke parts,uses the two-class iterative algorithm to classify the spacies, and find the bast split positions to recombined and merged those over-segmented components.Experimental results shows,this method is effective for the segmentation of online handwritten Chinese characters.