摘要
针对大多数草图识别算法笔画分组精确率低和用户适应性较差的问题,提出了一种基于混合特征的笔画分组方法,并在笔画的分组信息基础上构建了贝叶斯网络模型识别用户绘制的语义符号。该方法首先将用户绘制的笔画分组,每个分组代表一个独立的语义符号,然后提取分组的特征向量,最后通过贝叶斯网络模型推理出分组对应的语义符号。通过实验验证和数据分析,表明该方法具有良好的用户适应性、笔画分组能力及符号识别能力。
Because of a low accurative rate of sketch grouping and poor user- adaptive ability in most of sketch recognition algorithms, it proposes a method based on hybrid features of stroke used in sketches grouping. Based on the stroke sets obtained from first step, it builds a Bayesian network model for the user to identify the semantics of drawn sketches. This method groups the user - drawn sketches, makes each group represent a separated semantic symbol, and extracts feature vectors from the set of strokes. Finally, the semantics of the set corresponding to the symbol can be inferred through the Bayesian network model. The experiments verify that the method not only has a good ability in sketch grouping and symbol recognition, but also has a good adaptability of users, habits.
关键词
手绘草图
笔画分组
贝叶斯网络
符号识别
Hand- drawn Sketch
Sketch Grouping
Bayesian Network
Symbol Recognition
