关于三视图理解的研究

AN APPROACH TO THREE-VIEW DRAWING UNDER STANDING

本文论述了一种三视图的自动理解方法.概括为三部分即视觉信息处理部分、基元体分割部分及三维解释部分.本方法应用图象处理及模式识别技术,把输入图纸图象转换为易于机器处理的图形结构表示.利用机械制图中的某些特定规则及三视图间的对应关系,从组合体的三视图中提取出具有三维信息的基元体.以各个基元体的结合面为基准生成基元体间的位置维束关系,把各个基元体间的空间维束关系及它们的三维信息表示为一个具有两层结构的模型网络。实验表明该方法能较好地完成从图形描述向模型网络表示的转换.已能在积木世界为机器人的装配作业提供目标理解功能.

The approach includes three parts: (1) image information processing, (2) body primitives separating,(3) 3-D representation of engineering object. Part (1) symbolizes the image in figural form of loops of linesand curves suitable for higher level processing. Part (2) separates all body primitives which compose theobject in accordance with special expert rules in the field of engineering drawing and relations among itsthree views by heuristic inferring method. Part (3) represents the whole body on a two-level model net.This approach has been succesfully used by assembly robot in block world.