This page offers you some of my publications about "drawing interpretation". Feel free to download what ever you want. They are stored as encapsulated postscript compressed by gzip or WinZip. Gzip is a public domain compression tool from GNU. If you don't already have it, you will find it easily in the web (for various platforms).
J. M. Gloger, B. Pasternak, R. Sprengel, G. Gabrielides, N. Luth, M. Timmermann
1992, published paper (MVA'92), 4 pages, English text
Abstract:
The paper presents a system, which enables the conversion of paper-based technical drawings into a format suitable for CAD systems. The system is based on a commercial available scanning and vectorization unit. Since the vectorization results are far from being perfect, a postprocessing component to remedy some of the weak points have been created. The improved vectorization results permit a knowledge-based interpretation of the drawing content along with error detection and drawing redesign.
Boris Pasternak, Gabriel Gabrielides, Rainer Sprengel
1992, published paper (IEAAIE'92), 10 pages, English text
Abstract:
The WIZ-system is a prototype for interpretation of technical drawings. The paper introduces drawing interpretation as a knowledge processing task. Starting from a discussion about the necessary kinds of knowledge we present our frame-based knowledge structure and explain its implementation using a generic blackboard model, having easy adaptibility to new application domains in mind. The main part of the paper deals with the basic interpretation cycle which exploits a-priori knowledge about the drawing structure and associated geometrical constraints. This is discussed in detail for an example from the area of engineering drawings. The interpretation starts with the vectorization primitives and stops when physical objects (axles, cylinders, etc.) and non-physical dimensioning objects (dimension sets, metric arrows, etc.) are recognized. It is shown that the results of drawing interpretation can be used for domain-specific tasks, e.g. dimension checking or drawing redesign. Finally, a graph-based strategy for dimension checking is described in detail.
Boris Pasternak, Gabriel Gabrielides, Rainer Sprengel
1992, Memo, 32 pages, English text
Abstract:
This report describes the design and development of a prototype for the interpretation of technical drawings, called WIZ. The report introduces drawing interpretation as a knowledge processing task. At first, the state of the art in drawing analysis is examined. Then, starting from a discussion about the necessary kinds of knowledge to perform drawing interpretation, we present our frame-based knowledge structure and explain its implementation using a generic blackboard model. This allows easy adaptibility to new application domains. The main part of the report deals with the basic interpretation cycle which exploits a-priori knowledge about the drawing structure and associated geometrical constraints. This is discussed in detail for an example from the area of engineering drawings. The interpretation starts with the vectorization primitives and stops when physical objects (axles, cylinders, etc.) and non-physical dimensioning objects (dimension sets, metric arrows, etc.) are recognized. It is shown that the results of drawing interpretation can be used for domain specific tasks, e.g. dimension checking or drawing redesign. Finally, a graph-based strategy for dimension checking is described in detail.
1993, published paper (ICDAR'93), 6 pages, English text
Abstract:
This contribution deals with the problem of converting paper drawings into computer-based representations suitable for further use in CAD and CIM applications. While previous approaches have been able to solve this problem for certain application domains, the kernel system presented in this paper is adaptable to a wide variety of applications including engineering drawings and electronic circuit diagrams. The key elements of our approach are a knowledge-based interpretation process and a declarative specification language for defining meaningful graphical objects to be recognized in a drawing. The potential efficiency problems of a strictly declarative approach are overcome by automatically compiling graphic specifications into efficient procedures. The adaptability and efficiency of the kernel system are demonstrated using examples from several domains.
