摘要： 针对手绘草图识别算法大多采用限制用户绘制习惯来实现笔画分组的问题，提出一种基于贝叶斯网络的手绘草图识别算法。该算法将手绘草图识别中的笔画分组和符号识别统一为一个过程，用贝叶斯网络拓扑结构来表达草图结构信息。基于该网络，根据最大后验概率对连续输入的笔画进行动态最优分组，同时在线预测每组笔画的符号类别。实验结果表明，该方法是一种有效的在线递进式笔画分组和识别算法，在电路符号手绘识别中达到71.3%的过程识别率和85%的最终识别率。

关键词: 贝叶斯网络, 在线草图识别, 笔画分组, 符号识别

Abstract: To solve the limitation of restricting the user’s drawing style during the sketch grouping and recognition, a Bayesian network based sketch recognition algorithm is proposed. The algorithm combines the sketch grouping and the graphic symbol recognition into a unified procedure, which represents the sketch structure information as a Bayesian network. Based on the network, the growing sketches are grouped according to the maximum posterior probability, and each sketch group is recognized as a predefined symbol simultaneously. Experimental results show the effectiveness for the progressive sketch grouping and recognition, which has 71.3% procedure recognition rate and 85% final recognition rate.

Key words: Bayesian network, online sketch recognition, strokes grouping, symbol recognition

中图分类号: 

• TP391.41