• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Engineering design & automation >Knowledge-based semantics recovery module for mechanical engineering drawings
【24h】

Knowledge-based semantics recovery module for mechanical engineering drawings

机译：机械工程图的基于知识的语义恢复模块

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Most engineering drawings in active use today and most of those being produced are still in the hard copy format only. Digitization of a drawing creates an image of several million black and white pixels that represent the resolution of thescanning device and the quality of the original. However, this raster information is not suitable for CAD / CAF/CAM applications, which operate on data structures containing line segments, curves, surfaces, or higher level geometrical entities such asform features. Much research have been done in the past toward vector generation and character recognition from raster images. However, information recovered by mere vectorization and optical character recognition (OCR), such as those offered by existingcommercial systems, are too low level to be really useful in a CAD / CAF / CAM context. The process of semantics recovery after vectorization and OCR (i.e., converting the vectors into a set of meaningful drawing entities and associating them with thedrawing symbols) has gained little success. As such, the design knowledge embedded in engineering drawings fails to be extracted for downstream applications. In this article we describe a blackboard-architecture (BBA)based reasoning module for recoveringthree types of semantic information from the vector and character data: the topological relationships (i.e., incident and adjacent relationships) between the part geometric primitives, the composite drawing entities representing part annotationinformation, and the association between part annotation information and the corresponding part geometric primitives. Drawing entities of different levels of abstraction are organized in the blackboard data structure as a hierarchy consisting of sixlevels. The knowledge for interpreting the drawing entities is represented as a set of constraint nets. Each constraint net contains a "triggering node" used to invoke the constraint. Processing of a constraint is done by tracing the constraint net alongits nodes and links. A knowledge acquisition module is integrated with the reasoning module to learn new drawing patterns and add them to the knowledge sources. Because existing engineering drawing conversion systems lack the semantic recovery capacity,the proposed BBA based reasoning module addresses a major technology void in the current engineering drawing conversion process.
机译：今天，大多数正在使用的工程图以及正在制作的大多数工程图仍仅采用硬拷贝格式。图纸的数字化会产生数百万个黑白像素的图像，这些图像代表扫描设备的分辨率和原始图像的质量。但是，此栅格信息不适用于CAD / CAF / CAM应用程序，该应用程序在包含线段，曲线，曲面或更高级别的几何实体（例如形状特征）的数据结构上运行。过去，在矢量生成和光栅图像字符识别方面已经进行了许多研究。但是，仅由矢量化和光学字符识别（OCR）所恢复的信息（例如现有商业系统提供的信息）的级别太低，无法在CAD / CAF / CAM上下文中真正有用。向量化和OCR之后的语义恢复过程（即，将向量转换为一组有意义的绘图实体并将其与绘图符号相关联）几乎没有成功。因此，嵌入工程图的设计知识无法提取用于下游应用程序。在本文中，我们描述了一种基于黑板-架构（BBA）的推理模块，用于从矢量和字符数据中恢复三种类型的语义信息：零件几何图元之间的拓扑关系（即，入射关系和相邻关系），合成图形实体表示零件注释信息，以及零件注释信息与相应零件几何图元之间的关联。不同抽象级别的图形实体在黑板数据结构中组织为由六个级别组成的层次结构。用于解释图形实体的知识表示为一组约束网。每个约束网都包含一个用于触发约束的“触发节点”。通过沿着约束节点的节点和链接跟踪约束网络来完成约束的处理。知识获取模块与推理模块集成在一起，以学习新的绘图模式并将其添加到知识源中。由于现有的工程图转换系统缺乏语义恢复能力，因此提出的基于BBA的推理模块解决了当前工程图转换过程中的主要技术空白。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号