Sergey Ablameyko and Tony Pridmore. Machine Interpretation of Line Drawing Images. Technical Drawings, Maps and Diagrams.

Transcription

1 Sergey Ablameyko and Tony Pridmore Machine Interpretation of Line Drawing Images Technical Drawings, Maps and Diagrams i Springer

2 Sergey Ablameyko, PhD, DSc, Prof, FlEE, FIAPR, SMIEEE Institute of Engineering Cybernetics, Belarusian Academy of Sciences, Minsk, Belarus Tony Pridmore, BSc, PhD School of Computer Science and Information Technology, University of Nottingham, Nottingham, UK ISBN Springer-Verlag London Berlin Heidelberg British Library Cataloguing in Publication Data Ablameyko, Sergey Machine interpretation of line drawing images: technical drawing, maps and diagrams 1.0ptieal pattern recognition 2.Image processing - Digital techniques I.Titie II.Pridmore, Tony 621.3'67 ISBN Library of Congress Cataloging-in-Publication Data Ablameyko, Sergey, Machine interpretation of line drawing images: technical Drawings, maps and diagrams I Sergey Ablameyko and Tony Pridmore. p. em. Includes bibliographical references. ISBN-13: e-isbn-13: DOl: / Image proeessing--industrial applications. 2. Document imaging systems. 3. Optical scanners. 4. Optical pattern recognition. 1. Pridmore, Tony, II. Title. TA1637.A '7--de Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced, stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms of licences issued by the Copyright Licensing Agency. Enquiries concerning reproduction outside those terms should be sent to the publishers. Springer-Verlag London Limited 2000 Softcover reprint of the hardcover 1 5t edition 2000 The use of registered names, trademarks etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant laws and regulations and therefore free for general use. The publisher makes no representation, express or implied, with regard to the accuracy of the information contained in this hook and cannot accept any legal responsibility or liability for any errors or omissions that may be made. Typesetting: Camera ready by authors Printed on acid-free paper SPIN

3 Preface This book is concerned with the theory and practice of computer interpretation of images of line drawings. At present, many companies, agencies and individuals throughout the world regularly invest significant financial and human resource in the input of information into computers. Much of the information involved is presented in the form of line drawings. Manual input of those drawings to, for example, a Computer-aided Design (CAD) or Geographical Information System (GIS) is a laborious and time-consuming task. Line drawings are often large, containing many narrow, closely spaced line segments, each of which must be separately identified in the computer-based representation. Manual input generally involves tracing over the drawing with an appropriate pointing device, an operation that produces at best variable and often erroneous results. The availability of automatic and/or easily usable, interactive line drawing input systems would make this important process both easier and cheaper. The technological infrastructure needed to support the development of such systems is already in place. Equipment capable of efficiently scanning a document to produce a digital image is now widely available, as is software capable of basic image manipulation, storage and display. The extraction from the resulting image of the representations required by target applications such as CAD and GISs is, however, a complex problem that remains an active area of research and development. Line drawings arise in a wide variety of disciplines and situations. To provide a focus for discussion, we will deal primarily with the transformation of images of engineering drawings and maps into formats required by mechanical/manufacturing engineers and cartographers respectively. Many of the issues discussed and techniques described are, however, relevant to the interpretation of other types of line drawing. One might wish to consider, for example, images of floor plans, electrical schematics or data flow diagrams. As in these other domains, machine interpretation of maps and engineering drawings involves recognition of quite abstract graphical objects. Engineering drawings might be described in terms of, e.g. the outlines of physical objects, cross-hatched areas, centre lines and dimension sets. Entities of similar complexity (contour lines, roads, pipelines and cables, etc.) must be extracted from images of maps. Identification of these high level drawing constructs requires considerable knowledge of the type of drawing concerned; it is rare for someone

4 VI involved in manual drawing input to simply copy the document without giving at least some thought to its contents. Current commercial document input systems, however, only begin to exploit prior knowledge of drawing type and typically produce lower level representations than are really needed. Research results must be exploited, and further research is required, if useful interpretations are to be delivered at acceptable levels of speed and accuracy. In writing this book we have sought to gather together relevant information on the current state of line drawing interpretation and present it, along with the underlying theory, in a manner accessible to both academics and to the actual and potential developers of industrial document interpretation systems. The aim of the book is to show how line drawing interpretation systems can be developed. To this end we consider the problems to be faced (Chapterl), overview the technology underlying line drawing interpretation (Chapter 2) and consider the key operations (Chapters 3-7). To help the reader to realise the technology, detailed descriptions are given of selected methods, algorithms and data structures. To provide examples of the formation of complete interpretation systems from the available components we consider our own and our close colleagues' work in some detail. The map (Chapters 8 and 9) and engineering drawing (Chapter 10) interpretation systems developed by Prof. Ablameyko's group at Minsk, Belarus are discussed, as is the ANON system developed by Drs Steve Joseph and Tony Pridmore in Sheffield, England. Systems developed by others and/or applied in other domains are discussed more briefly, not because they are any the less impressive or important, but to allow us to exploit our own experiences and produce, hopefully, a more accurate and immediate text. Line drawing interpretation is a challenging area with enormous practical potential; we hope that our book can convey something of the spirit of the endeavour. Sergey Ablameyko Tony Pridmore Minsk Nottingham June 2000

5 Acknowledgements We would like to express our immense gratitude to the Belarusian and British colleagues with whom we have worked on line drawing interpretation systems and techniques. In particular, our colleagues from the Laboratory of Image Processing and Recognition, Institute of Engineering Cybernetics, Minsk, Belarus and the Department of Mechanical and Process Engineering, University of Sheffield, Sheffield, UK From Minsk, we particularly wish to thank V. Bereishik, O. Frantskevitch, M. Homenko, N. Paramonova, O. Patsko, V. Starovoitov, G. Aparin and D. Lagunovsky. Their support has allowed Prof. Ablameyko to develop several generations of line drawing image interpretation system over the last 15 years. Some sections of this book are drawn from papers previously written and published by Prof. Ablameyko in collaboration with these colleagues. Sections 5.6, 6.2.3, 6.3.3, 7.3.2, 7.6, 8.4, 9.1, 9.2, 9.4, and Chapter 10 benefit from contributions by V. Bereishik, O. Frantskevitch, N. Paramonova and M. Homenko. Dr. G. Aparin contributed to Section 8.6. The results presented in Section 9.3 were obtained in collaboration with Dr. V. Starovoitov. A. Kryuchkov developed the software described in Section 8.5. Finally, special thanks go to Prof. V. Tanaev for his permanent support of this research. In the UK we wish to express our particular thanks to Dr. Steve Joseph. Dr Joseph first introduced Tony Pridmore to the line drawing interpretation problem and Dr Pridmore enjoyed many discussions on this and other subjects during his time as Research Associate in Dr Joseph's laboratory. The ANON system was developed as part of an SERC ACME project headed by Dr Joseph and the description of ANON given here draws heavily upon papers written and published jointly by Drs Joseph and Pridmore. Many thanks are due for his permission to use this material. The initial idea for this book arose in the autumn of 1994, when Prof. Ablameyko visited Sheffield under a Royal Society Exchange Agreement. The Royal Society also supported the production of the text by awarding Tony Pridmore a Visiting Fellowship to the Institute of Engineering Cybernetics, Minsk for two weeks in the autumn of Without this support it is unlikely that the book would have been completed. Equally important is the personal support provided by our families; particularly Natasha, Andrei and Masha Ablameyko, and Lynn, Ellen, Joe and Amy Pridmore. Many thanks to all of you.

6 Contents 1. The Line Drawing Interpretation Problem Motivation Manual Input vs Document Scanning Raster-Based vs Vector Representations The Interpretation Problem Engineering Drawings and Maps Line Drawing Interpretation, Image Understanding and Pattern Recognition Current Line Drawing Interpretation Systems Commercial Systems Laboratory Systems The Line Drawing Interpretation Literature Components of a Line Drawing Interpretation System Design Criteria for Drawing Interpretation Systems Five Stages of Line Drawing Interpretation Intermediate and Target Representations Raster Representations Vector Representations Universal Drawing Entities Two-Dimensional Objects Three-Dimensional Shape and Semantics System Architectures and the Role of A Priori Knowledge Sequential Architectures Blackboard Architectures: The Expert System Approach Document Image Acquisition Scanning Devices Image Coding Image File Formats Binarisation A Taxonomy of Thresholding Techniques Document Image Statistics Binarising Line Drawings Point-Dependent Threshold Selection... 64

7 x Region-Dependent Threshold Selection Image Partitioning Binary Image Processing and the Raster to Vector Transformation Raster to Vector Conversion Some Defmitions The Distance Transform Mathematical Morphology Reducing Noise in Binary Images Reducing Noise in Binary Images of Line Drawings Analysis of Connected Components Nomenclature Contouring Goals Classification of Contouring Techniques A Contouring Algorithm Skeletonisation Motivation Approaches To Thinning A Thinning Algorithm The Medial Axis Transform Removing Noise from Thinned Data Grey Level Skeletonisation Vectorisation Approaches To Vectorisation Global Vectorisation Methods Iterative Methods Feature-Based Techniques Hybrid Approaches Local Vectorisation Methods Line Following Methods Raster Scan Techniques The Hough Transform Direct Vectorisation A Vector Database Removing Noise from the Vector Model Alternative Raster to Vector Technologies Interpreting Images of Maps Introduction System Overview Map Interpretation Principles A Classification of Map Entities Line Objects Area Objects

8 8.4.3 Symbols Interactive Map Interpretation Automated Image Interpretation under Operator ControL Operator-Supplied ConteXt Output Formats Quality Issues Basic Concepts Maintaining Quality during Interpretation Recognising Cartographic Objects Recognising Isolines Overview Theoretical Background An Isoline Recognition Procedure Recognising Roads Overview Interpreting the Road Layer Extracting Roads and Correlated Objects from the Black Layer Recognising Texture and Area Objects Overview Texture Border Definitions Recognising Symbols Recovering Engineering Drawing Entities from Vector Data Design Principles and System Architecture Vectorisation and Entity Recognition Processes Extracting Arcs and Straight Lines loa Recognising Crosshatched Areas Recognising Dimensions Detecting Blocks Knowledge-Directed Interpretation of Engineering Drawings An Image Understanding Approach Drawing Entities as Schemata Image Analysis Facilities The ANON Architecture Control Issues Entity Extraction: Chained Lines Top-Down and Bottom-Up Control Performance Scene Formation Searching Image and Memory Coincidence Links Partial Interpretations and Schema Fusion ANON in Context Discussion XI

9 XII 12. Current Issues and Future Developments Higher-Level and 3D Representations Resolving Inconsistency Editing and Parameterisation D Reconstruction Exploiting Domain Knowledge The Role of the Operator Performance Measures Topics for Future Development References Index