Abstract: The current election systems or ballot recognition algorithms simplify layout understanding, and can not be applied to different kinds of ballots. Based on layout segmentation, this paper presents a description method of layout structure based on undirected graph, which makes location correctly and batch counting fast according to various ballot forms. Experiment shows that the method is agile, accurate, general and fast.

Key words: undirected graph, depth-first traversal, ballot, layout understanding

摘要： 目前的选举系统或选票识别算法都简化了选票的版面理解，无法适应不同类型选票的准确定位、识别。该文在版面分割的基础上，提出一种基于无向图的选票版面结构描述方法，采用改进的深度优先遍历算法实现多种类型选票的准确定位和批量快速统计。实验证明，该方法具有选票设计灵活、精度高、通用性好、处理速度快等特点。

关键词: 无向图, 深度优先遍历, 选票, 版面理解

CLC Number: 

• TP391.41