基于语法描述语言的在线手绘流程图识别  

On-line Handwritten Flowchart Recognition Based on Grammar Description Language

作　　者：陈泉[1,2] 施大鹏[1,2] 冯桂焕[1,2] 赵小燕[1,2] 骆斌[1,2] 

机构地区：[1]南京大学软件学院,南京210093 [2]计算机软件新技术国家重点实验室,南京210093

出　　处：《计算机科学》2015年第B11期113-118,共6页Computer Science

基　　金：国家自然科学基金(61100109)资助

摘　　要：提出了一种基于语法描述语言来对草图的结构和组成符号进行识别的方法。该方法将整个识别过程分为3步:首先,使用具有空间和时序约束的动态规划对笔画序列进行自由组合;然后通过神经网络分类器对笔画组合进行识别,生成候选符号集;最后,基于流程图构成的语法规则,对候选符号进行筛选,最终通过语法解析获得识别结果。在FCinkML数据上对该方法进行验证,结果表明了该方法的有效性。This paper proposed a sketch recognition approach based on a grammar description language. The whole recognition process is divided into three steps. Firstly, dynamic programming is used to group strokes into stroke combinations under some spatial and temporal constraints. Secondly, a neural network is employed to classify the candidate symbols. Lastly, the candidate symbols are filtered with the constituent grammar rules, and a grammar parser is utilized to achieve the recognition results. This method has been applied to a freely available database FCinkML. The results demonstrate the effectiveness and efficiency of our approach.

关 键 词：草图理解 手绘流程图 语法描述语言 神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]