A database capture system for mechanical drawings using an efficient multidimensional graphical data structure

A novel type of system for understanding mechanical drawings, named the AI-MUDAMS Recognizer, is proposed and discussed. In the system, contour vector data and core vector data for objects in the drawing image are first extracted and stored in a multidimensional graphical database. All the geometrical operations required for feature extraction and structural understanding of mechanical drawings are performed very efficiently and flexibly using only the vector data (rather than image data) in the database. Rule-based artificial intelligence techniques are used for structural understandings. Experimental results confirm the high speed of the systems and its ability to understand mechanical drawings