摘要
提出一种统计特征点网格分布的表格图像识别方法 ,该方法以表格框线间的交叉点类型作为表格分类的主要结构特征 ,把表格图像外接矩形区域归一化为N×N的网格 ,并统计每一网格内各种类型特征点的分布情况 ,由此形成的N×N个向量作为表格识别的特征向量 .采用了类似度的方法作为表格分类的判别准则 ,将未知表格类型的特征向量与预先经过学习建立的表格模板库中的标准特征向量进行相似性度量 ,取其类似度最高的模板类型作为识别结果 .实验表明该方法可行、高效 .
A table form document recognizing method is proposed for counting the feature points in the mesh of the table. The cross point of the table line is used as its structure feature, with the table area divided into N×N meshes the distribution of all kinds of feature points are analyzed to obtain the N×N dimension vectors which can be used as the feature vectors of form recognition. By adopting the similarity method, the vectors of the unknown type of table are compared with the standard vectors of the table template, with the type of the highest similarity as the result of recognition. It has been proved by experiment that the method is both feasible and highly efficient.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2002年第9期60-63,共4页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
