选票版面结构识别相关技术  

Relevant technology of ballot structure identification

作　　者：代威 官磊 韩啸 DAI Wei;GUAN Lei;HAN Xiao(Chengdu Institute of Computer Application,Chinese Academy of Sciences,Chengdu Sichuan 610041,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院成都计算机应用研究所,成都610041 [2]中国科学院大学,北京100049

出　　处：《计算机应用》2019年第S02期242-246,共5页journal of Computer Applications

摘　　要：针对目前选票处理过程选票物理结构识别和逻辑结构识别存在的问题。提出一种在改进的游程线框检测基础上进行逻辑符号识别的方法。首先对扫描图像进行二值化、倾斜矫正等预处理,接着利用同线游程遍历线框,再通过设定的线段重合度阈值合并相关度高的线段,最后根据定义的垂直连接线判定式找到对应线框。在此之后对线框内字符逻辑结构进行识别,利用双向游程原理设定符号游程,并对符号区域特征优化后,结合游程坐标约束集判定符号所属类别。通过实验对比标记块算法,在保证时间效率的同时,改进算法对符号的处理准确度达到99%,能满足实际选举需求。For current ballot processing process,there are problems in the physical structure identification and logical structure identification of the ballot.A method for logical symbol recognition based on improved run-line frame detection was proposed.Firstly,a scanned image was preprocessed by binarization,tilt correction,etc.,and then the line frames were traversed by the same line run,and then the line segments with high correlation were merged by the set line segment coincidence threshold.Finally,the corresponding wireframes were found according to the defined vertical connection line judgment formula.After that,the logical structure of the characters in the wireframe were identified.The symbol runs were set by using the two-way run principle,and then the symbol region features were optimized and combined with the run coordinate constraint set to determine the category to which the symbol belongs.By comparing the marker block algorithm with experiments,while improving the time efficiency,the improved algorithm can process the symbols with an accuracy of 99%,which can meet the actual election requirements.

关 键 词：电子选举 选票 符号识别 版面理解 表格游程 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]