一种基于图元结构关系的电气草图符号识别方法

摘要

Aimirg at the problem of difficult to recognize and understand the hand-drawn symbols, because of irregular and fuzzy, the features of hand-drawn electrical symbols, stroke segmentation of the symbols, the identification of the primitive type, structural relations between primitives, the types of the similarity between symbols and other aspects were studied, and a new recognition method for hand-drawn electrical symbols based on structural relationship between primitives was proposed. First of all, hand-drawn electrical symbols were segmen-ted into several primitives based on their characteristics. Then attributed relational graph was used to describe the attribution of the primitives and the structural relationship between primitives in hand-drawn electrical symbols. And a method for calculating the similarity of hand-drawn electrical symbols matching based on attributed relational graph was proposed. Finally, through calculating the similarity between the unknown symbol and the symbol in template library, the unknown hand-drawn electrical symbol was recognized successfully. The results indi-cate that the proposed method can be used to construct an effective, trainable, multiline character recognizer, which is not affected by the direction, the size and the drawing order of the symbol.%针对手绘草图符号不规则以及模糊的特点造成对其识别难度较高的问题,对手绘电气草图符号的特点、符号的笔划分割、图元类型的判别、图元间的结构关系、符号间相似度的类型等方面进行了研究,提出了一种基于图元结构关系的手绘电气草图符号识别方法.根据手绘电气草图符号的特点,首先对手绘电气草图符号进行了笔画分割,将其分解成了若干基本图元;然后用属性关系图描述出了手绘电气草图符号单个图元的信息以及图元间的结构关系,提出了一种基于属性关系图的电气草图符号匹配相似度计算方法;最后通过计算待识别符号与模板库符号之间的相似度,达到了手绘电气草图符号识别的目的.研究结果表明:可以构建一个有效的、可训练的、多线条的符号识别器,并且不受符号方向、大小和笔画顺序的影响,能够取得良好的识别效果.