Lecture Notes in Geoinformation and CartographySeries Editors: William Cartwright, Georg Gartner.Liqiu Meng,

Michael Peterson

Alias Abdul-Rahman • Sisi Zlatanova·Volker Coors (Eds.)

Innovations in 3D GeoInformation Systems

With 438 Figures

~ Springer

Editors:

Dr. Alias Abdul-RahmanDepartment of Geoinformatics,Faculty of GeoinformationScience and EngineeringUniversiti Teknologi Malaysia81310 Skudai, Iohor, MalaysiaEmail: alias@fksg.utm.my

Dr. Sisi ZlatanovaSection GISt, OTBDelft University of TechnologyPostbus 50302600 GA Delft, The NetherlandsEmail: s.zlatanova@otb.tudelft.nl

Dr. Volker CoorsGeoinformaticsFaculty Geomatics, Mathematicsand Computer ScienceUniversity of Applied SciencesSchellingstr. 2470174 Stuttgart, GermanyEmail: volker.coors@hft-stuttgart.de

ISBN 10ISBN 13

3-540-36997-x Springer Berlin Heidelberg New York978-3-540-36997-4 Springer Berlin Heidelberg New York

Library of Congress Control Number: 2006930254

This work is subject to copyright. All ri~hts are reserved, whether the whole or l?,art of thematerial is concerned, specifically the nghts of translation, reprinting, reuse of Illustra­tions, recitation, broadcasting, reproduction on microfilm or in any other way, and stor­age in data banks. Duplication of this publication or parts thereof is permitted only underthe provisions of the German Copyright Law of September 9, 1965, in its current version,and permission for use must always Deobtained from Springer-Verlag. Violations are li­able to prosecution under the German Copyright Law.

Springer is a part of Springer Science+ Business Mediasprin&eronline.com© Springer-Verlag Berlin Heidelberg 2006

The use of general descriptive names, 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 protective laws and regulations and therefore free for general use.

Cover design: E. Kirchner, HeidelbergProduction: A. OelschlagerTypesetting: Camera-ready by the Editors

Printed on acid-free paper 30/2132/AO 54321

Foreword

Three-dimensional (3D) geoinformation has been the subject withingeoinformation (GIS) community and other related professionals sinceearly 90's. Recently the subject is getting more attention when users de­mand better information of the phenomena and of the surrounding objects.A lot of efforts and time have been invested for this kind of 3D geoinfor­mation solution but operationally such system is still hardly available inmass market.

For this reason the Department of Geoinformatics, Faculty of Geoin­formation Science and Engineering, Universiti Teknologi Malaysia tookthe initiative to organize the first International Workshop on 3D Geoin­formation held in South East Asia, held on 7-8 August, 2006, Kuala Lum­pur, Malaysia. The workshop has covered fundamental aspects of 3Dgeoinformation as well as 3D data collection (LIDAR and other means),3D spatial data modeling, 3D databases, 3D spatial analysis, 3D visualiza­tion, and some applications in 3D geoinformation.

This workshop has gathered researchers, system developers, practitio­ners, and end users from 21 countries, coming from 5 different continents.They have the great opportunity to discuss new research development,ideas, theories, and possible applications in this very specialized but fastemerging field.

This event has been supported by several other universities amongstwhich TU Delft, The Netherlands and University of Applied Sciences,Stuttgart, Germany; government agencies such as Department of Surveyand Mapping Malaysia (JUPEM); professional unions such as ISPRS (In­ternational Society of Photogrammetry and Remote Sensing) and LandSurveyors Board of Peninsular Malaysia (LIT); and the international soft­ware vendor like Autodesk. Their commitment and support gives usstrength and surely motivates all of us in this very specialized subject forbetter solutions and services.

Goal and objectives

The workshop emphasizes on the third dimension of geographical informa­tion science (GISc.). It is meant to be an interdisciplinary forum for lead­ing researchers in related areas to present the latest 3D developments and3D applications, to discuss cutting-edge 3D technology, to exchange re-

vi

search ideas and to promote international collaboration in this field. Theconference will concentrate on the following areas:

• Data collection and modeling: advanced approaches for 3D data collec­tion, 3D re-construction and methods for representation.

• Data Management: topological, geometrical and network models formaintaining of 3D geo-information.

• Data analysis and visualization: Frameworks for representing 3D spatialrelationships, 3D spatial analysis and algorithms for navigation, interpo­lation, etc. Advanced VR, AR, MR visualization, 3D visualization onmobile devices

• 3D applications: City models, cadastre, LBS, etc.

The Paper Selection Process

The International Workshop on 3D Geoinformation is a refereed work­shop. We received more than 70 technical papers. After a careful selectionwe have admitted 51 papers, which are included in this book. The selectedpapers were divided into two categories, i.e, oral and poster presentationsas indicated in this book.

Acknowledgement

We thank to our major sponsors like Autodesk, Land Surveyors Board ofPeninsular Malaysia (LJT), and Department of Survey and Mapping Ma­laysia for their excellent supports and contributions. Thanks also go tosoftware and geo services vendors and companies for their participation inthe exhibition. Finally, we highly appreciate the efforts, helps and contri­butions from committee members (local and international), and other indi­viduals.

Alias Abdul Rahman,Sisi Zlatanova,Volker Coors.

10 July, 2006

Table of Contents

Keynotes

3D Geometries in Spatial DBMS 1Sisi Zlatanova (The Netherlands)

A Web 3D Service for Navigation Applications 15Jorg Haist, Thorsten Reitz, Volker Coors (Germany)

3D Spatial Data Acquisition - LIDAR and DigitalPhotogrammetry

Oral Contributions

Integration of Photogrammetric and LIDAR Data in a 29Multi-Primitive Triangulation EnvironmentA. F. Habib, S. Shin, C. Kim, and Muhannad Al-Durgham (Canada andSouth Korea)

LIDAR-Aided True Orthophoto and DBM 47Generation SystemA. F. Habib and C. Kim (Canada)

Surface Matching Strategy for Quality Control of 67LIDARDataA. F. Habib and Rita W. Cheng (Canada)

On-line Integration of Photogrammetry and GIS 85to Generate Fully Structured Data for GISHamid Ebadi, Farshid Famood Ahmadi (Iran)

x

3D Spatial Data Modelling and Representation

Oral Contributions

3D Integral Modelling for City Surface and 95SubsurfaceWang Yanbing, Wu Lixin, Shi Wenzhong and Li Xiaojuan (China, andHong Kong)

Spatial Object Structure for Handling 3D Geodata 107inGRIFINOREric Kjems and Jan Kolar (Denmark)

The Study and Applications of Object-Oriented 119Hyper-Graph Spatio- Temporal Reasoning ModelLuo Jing, Cui Weihong and Niu Zhenguo (China)

Using 3D Fuzzy Topological Relationships for 129Checking of Spatial Relations between Dynamic Air PollutionCloud and City Population DensityRoozbeh Shad, Mohamad Reza Malek, Mohammed Saadi Mesgari, andAlireza Vafaeinezhad (Iran)

3D Modelling Moving Objects Under Uncertainty Conditions 139Tala Shokri, M. R. Delavar, M. R., Malek, A. U. Frank, and G. Navratil(Iran, and Austria)

Research on a Feature Based Spatio-Temporal Data Model 151Weihong Cui, Wenzhong Shi, Xiaojuan Li, Luojing, and Zhengguo Niu(China and Hong Kong)

OD Feature in 3D Planar Polygon Testing for 1693D Spatial AnalysisChen Tet Khuan and Alias Abdul Rahman (Malaysia)

Definition of the 3D Content and Geometric Level 185of Congruence of Numeric CartographyR. Brumana, F. Fassi, and F. Prandi (Italy)

xi

3D Multi-scale Modelling of the Interior of the 195Real Villa of Monza (ITALY)C. Achille and F. Fassi (Italy)

3D GIS Frameworks

Oral Contributions

On the Road to 3D Geographic Systems: 207Important Aspects of Global Model-Mapping TechnologyJan Kolar (Denmark)

Cristage: A 3D GIS with a Logical Crystallographic ............•..... 225Layer to Enable Complex AnalysisB. Poupeau and Olivier Bonin (France)

The Democratizing Potential of Geographic Exploration 235Systems (GES) Through the Example of GRIFINORLars Bodum and Marie Jaegly (Denmark)

The Integration Methods of 3D GIS and 3D CAD 245Li Juan, Tor Yam Khoon, and Zhu Qing (China and Singapore)

3D Navigation for 3D GIS - Initial Requirements 259Ivin Amri Musliman, Alias Abdul Rahman, and Volker Coors (Malaysiaand Germany)

Web-based GIS-Transportation Framework Data 269Services Using GML, SVG and X3DHak-Hoon Kim and Kiwon Lee (South Korea)

3D Geo-Database Implementation Using ..........•...•.................. 279Craniofacial Geometric Morphometries Database SystemDeni Suwardi and Halim Setan (Malaysia)

GIS-based Multidimensional Approach for Modelling .........•....• 295Infrastructure InterdependencyRifaat Abdalla, Haris Ali, and Vincent Tao (Canada)

xii

Conception of a 3D Geodata Web Service for the 307Suppport of Indoor Navigation with GNSSStephen Mas, Wolfgang Reinhardt and Fei Wang (Germany)

3D Objects Reconstruction

Oral Contributions

Reconstruction of 3D Model based on Laser Scanning 317Lu-Xingchang and Liu-Xianlin (China)

Automatic Generation of Pseudo Continous LoDs 333for 3D Polyhedral Building ModelJiann-Yeou Rau, Liang-Chien Chen, Fuan Tsai, Kuo-Hsiao and Wei-ChenHsu (Taiwan)

Reconstruction of Complex Buildings Using 345LIDAR and 2D MapsTee-Ann Teo, Jiann-Yeou Rau, Liang-Chien Chen, lin-King Liu, and Wei­Chen Hsu (Taiwan)

Building Reconstruction - Outside and In 355Christopher Gold, Rebecca Tse, and Hugo Ledox (United Kingdom)

Skeletonization of Laser-Scanned Trees in the 3D 371Raster DomainBen Gorte (The Netherlands)

Automated 3D Modelling of Buildings in Suburban 381Areas based on Integration of Image and Height DataKourosh Khoshelham (Iran)

Automatically Extracting 3D Models and 395Network Analysis for IndoorsIsmail R. Karas, Fatmagul Batuk,Abdullah E. Akay, and Ibrahim Ba:(Turkey)

xiii

3D City Modelling

Oral Contributions

Improving the Realism of Existing 3D City Models 405Martin Kada, Norbet Haala, and Susanne Backer (Germany)

Different Quality Level Processes and Products 417for Ground-based 3D City and Road ModelingBahman Soheilian, Olivier Toumaire, Lionel Penard, Nicolas Paparoditis(France)

Texture Generation and Mapping using Video 429Sequences for 3D Building ModelsFuan Tsai, Cheng-Hsuan Chen, Jin-Kim Liu, and Kuo-Hsing Hsiao (Tai­wan)

Design and Implementation of Mobile 3D City 439Landscape AuthoringlRendering SystemSeung-Yub Kim and Kiwon Lee (South Korea)

Macro to Micro Archeological Documentation: 447Building a 3D GIS Model for Jerash City and the Artemis TempleNedal Al-Hanbali, Omar Al Bayari, Bassam Saleh, Husam Almasri, Em­manuel Baltsavias (Jordan and Switzerland)

Building 3D GIS Modelling Applications in Jordan: 469Methodology and Implementation AspectsNedal Al-Hanbali, Eyad Fadda, and Sameeh Rawashdeh (Jordan)

3D Mapping, Cadastre and Utility

Oral Contributions

Moving Towards 3D - From a National Mapping Agency 491Dave Capstick and Guy Heathcote (United Kingdom)

xiv

An Approach for 3D Visualization of Pipelines 501Y. Du and S. Zlatanova (China and The Netherlands)

Developing Malaysian 3D Cadastre System - 519Preliminary FindingsMuhammad Imzan Hassan, Alias Abdul Rahman, and Jantin Stoter (Ma­laysia and The Netherlands)

Developing 3D Registration System for 3D Cadastre 535Mohd HasifAhmad Nasruddin and Alias Abdul Rahman (Malaysia)

3D Visualization

Oral Contributions

Volumetric Spatio Temporal Data Model 547Mohd Shafry Mohd Rahim, A. R. Sharif, S. Mansor, A. Rodzi Mahmud, andMohammad Ashari Alias (Malaysia)

Use of 3D Visualization in Natural Disaster Risk ........•.............. 557Assessment for Urban AreasS. Kemec and H. S. Duzgun (Turkey)

Development and Design of 3D Virtual Laboratory 567for Chemistry Subject based on Constructivism-Cognitivism­Contextual ApproachHajah Norasiken Bakar, and Halimah Badioze Zaman (Malaysia)

The 3D Fusion and Visualization of Phototopographic Data 589Yang Ming Hui, Ren Wei Chun, Guan Hong Wei, and Wang Kai (China)

Integrating a Computational Fluid Dynamics 599Simulation and Visualization with a 3D Virtual Walkthrough -A Case Study of PutrajayaPuteri Shireen Jahnkassim, Maisarah Ali, NoorHanita A. Majid, and Man­sor Ibrahim (Malaysia)

xv

A Geospatial Approach to Managing Public 615Housing on SuperlotsJack Barton and Jim Plume (Australia)

3D Visualization and Virtual Reality for 629Cultural Heritage DiagnosticL. CoLizzi, F. D Pascalis, and F. Fassi (Italy)

3D Terrain Modeling and Digital Orthophoto Generation

Oral Contributions

True Orthophoto Generation from High 641Resolution Satellite ImageryAyman F. Habib, K. I., Bang, C. J. Kim, and S. W. Shin (Canada)

Development of Country Mosaic using IRS-WIFS Data 657Ch. Venkateswara Rao, P. Sathyanarayana, D. S. Jain, A. S. Manjunath,and K. M. M. Rao (India)

Digital Terrain Models Derived from SRTM 673Data and KrigingT. Bernardes, I. Gontijo, H. Andrade, T. G. C. Vieira, and H. M. R. Alves(Brazil)

The St Mark's Basilica Pavement: 683The Digital Orthophoto 3D Realization to The Real Scale 1:1 forThe Modelling and The Conservative RestorationL. Fregonese, C. C. Monti, G. Monti, and L. Taffurelli (Italy)

xvi

Poster Contributions

The Application of GIS in Maritime 695Boundary DelimitationI Made Andi Arsana, Chris Rizos, Clive Schofield (Indonesia and Austra­lia)

Integration of GIS and Digital Photogrammetry 721in Building Space AnalysisMokhtar Azizi Mohd Din and M. Yazi: Ahmad (Malaysia)

An Integration of Digital Photogrammetry 737and GIS for Historical Building DocumentationSeyed Youse! Sadjadi (United Kingdom)

Reconstruction of Three Dimensional 749Ocean Bathymetry Using Polarised TOPSAR DataMaged Marghany and Mazlan Hashim (Malaysia)

Author Index

A. U. Frank, 139A.F. Habib, 67A.S.Manjunath, 657Alias Abdul Rahman, 259, 519,535,

169Alireza Vafaeinezhad, 129Ayman Habib, 47B. Poupeau, 225Bahman Soheilian, 417Ben Gorte, 371C. C. Monti, 683Ch.Venkateswara Rao, 657Changjae Kim, 47Chen Tet Khuan, 169Cheng-Hsuan Chen, 429D.S,Jain, 657Dave Capstick, 491Deni Suwardhi, 279Eric Kjems, 107Eyad Fadda, 469F. De Pascalis, 629F. Fassi, 185,629F. Prandi, 185Fuan Tsai, 333, 429G. Monti, 683G. Navratil, 139Guan Hong Wei, 589Guy Heathcote, 491H. Andrade, 673H. M. R. Alves, 673Halim Setan, 279Halimah Badioze Zaman, 567Harris Ali, 295I. Gontijo, 673Ivin Amri Musliman, 259Jack Barton, 615Jan Kolar, 107Jan Kohli" 207Jantien Stoter, 519Jiann-Yeou Rau, 333, 345Jim Plume, 615Jin-Kim Liu, 429Jin-King Liu, 345

xvii

Jorg Haist, 15K.M.M.Rao, 657Kourosh Khoshelham, 381Kuo-Hsin Hsiao, 333Kuo-Hsing Hsiao, 429L. Colizzi, 629L. Fregonese, 683L. Taffurelli, 683LIJuan, 245Li Xiaojuan, 95Liang-Chien Chen, 333, 345Lionel Penard, 417M. R. Delavar, 139M. R. Malek, 139Maged Marghany, 749Mazlan Hashim, 749Mohamad Reza Malek, 129Mohammad Saadi Mesgari, 129Mohammed Yaziz Ahmad, 721Mohd Hasif Ahmad Nasruddin, 535Mokhtar Azizi Mohd Din, 721Muhammad Irnzan Hassan, 519Nedal AI-Hanbali, 469Nicolas Paparoditis, 417Norasiken Bakar, 567O. Bonin, 225Olivier Toumaire, 417P.Sathyanarayana, 657R. Brumana, 185R.W.T. Cheng, 67Ren Wei Chun, 589Rifaat Abdalla, 295Roozbeh Shad, 129Sameeh Rawashdeh, 469Seyed Yousef Sadjadi, 737ShiWenzhong,95Shokri, T, 139Sisi Zlatanova, 1T. Bemardes, 673T. G. C. Vieira, 673Tee-Ann Teo, 345Thorsten Reitz, 15TOR Yam Khoon, 245

xviii

Vincent Tao, 295Volker Coors, 15,259Wang Kai, 589Wang Yanbing, 95

Wei-Chen Hsu, 333, 345WuLixin, 95Yang Ming Hui, 589ZHU Qing, 245

xix

Program committee

ChairAlias Abdul Rahman, Universiti Teknologi Malaysia

Co-chairSisi Zlatanova, Delft University of Technology, The Netherlands

MembersChristopher Gold, University of Glamorgan, United KingdomDaniel Holweg, Fraunhofer Institute, GermanyHalim Setan, Universiti Teknologi MalaysiaJae-Hong Yom, Sejong University, South KoreaJantien Stoter, lTC, The NetherlandsJiyeong Lee, University of North Carolina at Charlotte, United StatesLars Bodum, Aalborg University, DenmarkLixin Wu, China University of Mining and Technologies, Beijing, China.Martin Kada, University of Stuttgart, GermanyMilan Konechny, University of Masaryk, Brno, Czech RepublicMichael Goodchild, University of California at Santa Barbara, UnitedStatesMichael Hahn, University of Applied Sciences Stuttgart, GermanyMonika Sester, University of Hannover, GermanyMorakot Pilouk, ESRI, California, United StatesNorbert Haala, University of Stuttgart, GermanyQingquan Li, Wuhan University, ChinaRoland Billen, Universite de Liege, BelgiumRyosuke Shibasaki, University of Tokyo, JapanSylvie Servigne, INSA-Lyon, FranceTaher Buyong, International Islamic Univ., MalaysiaThomas Kolbe, University of Bonn, GermanyVincent Tao, York University, Toronto, CanadaVolker Coors, University of Applied Sciences Stuttgart, GermanyWenzhong Shi, The Hong Kong Polytechnic University, Hong Kong

Local Organizing Committee

The Local Organizing Committee consists of staff members of the De­partment of Geoinformatics, Faculty of Geoinformation Science and Engi­neering, Universiti Teknologi Malaysia (UTM) , and the Department ofSurvey and Mapping, Malaysia (JUPEM)

xx

Abd Razak Abu Bakar, UTMAhmad Fauzi Nordin, mPEMAlias Abdul Rahman, UTMHalim Setan, UTMIyin Amri Musliman, UTMMohamad Nor Said, UTMMuhammad Irnzan Hassan, UTMNorkhair Ibrahim, UTMShahabudin Amerudin, UTMZamri Ismail, UTM

3D Geometries in Spatial DBMS

Sisi Zlatanova

GISt, Delft University of Technology, The Netherlandss.zlatanova@otb.tudelft.nl

Abstract

Database management systems (DBMS) have significantly changed in thelast several areas. From a system dealing with management of administra­tive data they have involved to a spatial DBMS providing spatial datatypes, spatial indexing and extended spatial functionality. MainstreamDBMS support mostly a geometry model but the first natively supportedtopology model is already a fact. Although the provided functionality islimited to the second dimension, various options exist for management ofthree-dimensional data. This paper discusses some of them, present currentresearch and outline directions for further extended management of 3Ddata. The paper is limited to the geometry model, i.e. topology issues arenot covered.

Evolving to Spatial DBMS

DBMS have been traditionally used to handle large volumes of data andto ensure the logical consistency and integrity of data, which also have be­come major requirements in handling spatial data. For years, spatial dataused to be organized in dual architectures consisting of separated datamanagement for administrative data in a Relational DBMS (RDBMS) andspatial data in a GIS. This is to say spatial data has been managed in singlefiles using proprietary formats. This approach could easily result in incon­sistency of data, e.g. when deleting a record in the database no mechanismexists to check the corresponding spatial counterpart. A solution to prob­lems of dual architecture was a layered architecture, in which all data ismaintained in a single RDBMS. Since spatial data types were at that time

2 Sisi Zlatanova

not supported at DBMS level, knowledge about spatial data types wasmaintained in middle ware (Vijlbrief and van Oosterom, 1992).

Presently, most mainstream DBMS offer spatial data types and spatialfunctions usually in an object-relational spatial extend to RDBMS (Zla­tanova and Stoter, 2006). Storing spatial data and performing spatialanalysis can be completed with SQL queries. Additionally, integrated que­ries on both spatial and non-spatial parts of features can be executed at da­tabase level. The spatial data types and spatial operations reflect only sim­ple two-dimensional features, though embedded in 3D space. This supportof 3D/4D coordinates allows for alternatives in management of 3D fea­tures.This paper elaborates on current possibilities of DBMS to maintain 3Ddata. The next section discusses management and visualisation of volumet­ric objects, 3D lines and 3D points. Then the paper reports on prototypeimplementations of new data types completed at section GISt, Delft Uni­versity of Technology, The Netherlands. Standardization activities withinOpen Geospatial Consortium are briefly outlined with respect to recentnew initiatives. Last section concludes on demands and expectations to the3D geometry model.

3D data in the DBMS within current techniques

Providing the spatial extend, DBMS have immediately been challenged bythe third dimension. A number of experiments were performed by severalresearchers to investigate possibilities to store, query and visualize featureswith their 3D coordinates (Arens et al 2006, Stoter and Zlatanova 2003,Pu, 2005, Zlatanova et al 2002, Zlatanova et al 2004). The good news is:3D data can be organized in DBMS, retrieved and rendered by front-endapplications. However, there is also a bad news: since no 3D data type iscurrently supported by any DBMS, the user remains self responsible forthe validation of the objects as well as for implementing true 3D function­ality.

These conclusions, with small variations, are consistent for all main­stream DBMS: Oracle, IBM DB2, Informix, Ingres, PostGIS and MySQL.All of them offer 2D data types (basically point, line, polygon) but support3D/4D coordinates (except Ingres, which is 2D) and offer a large numberof functions more or less compliant with the Open Geospatial Consortium(OGC) standards (see bellow). Most of the functions are only 2D (exceptPostGIS, which supports few 3D operations), i.e. although not reportingerror, they omit the z-coordinate. A bit frustrating is the implementations

3D Geometries in Spatial DBMS 3

of spatial functions and operations : it varies with the DBMS. The state­ment: select c from b where a < 100, where c,a are numerical data type,can be executed in every DBMS. However if c is a spatial data type and ais a given distance, the SQL statement becomes depended on the specificimplementation. In some cases (e.g. Oracle Spatial), even the names of thespatial data types are not that apparent. Oracle Spatial has one complexdata type sdo jgeometry composed of several parameters indicating typegeometry, dimension, and an array with the x,y,z coordinates. At presentthe geometry model of Oracle Spatial is the most supported by GIS andArchitecture, Engineering and Construction (ABC) Systems, but there is astrong tendency for changes (e.g. PostGIS is supported by GRASS andESRI). The text below discusses possible organisation of the 3D real-worldfeatures (volumetric, line, point) in current DBMS. Note, the presented ap­proaches are not dependent on the DBMS.

Fig. 1. Visualisation of buildings and surface, represented by simple polygons inOracle Spatial

3D volumetric objects

Most discussed 3D features are volumetric objects, which can be used formodelling of man-made objects, such as buildings, and natural objects,such as geological formations. To have those managed in the database, the

4 Sisi Zlatanova

user can choose between: 1) using DBMS data types polygon and multi­polygon or 2) creating a used-defined data type. The three candidates for asimple volumetric object are polyhedron, triangulated polyhedron and tet­rahedron (see for definitions next section) and all three can be easily real­ized with provided data types. Moreover, there is no practical difference inthe implementation of the polyhedron and triangulated polyhedron, since aseparate triangle data type does not exist. Tetrahedron would allow for abit simpler representation since it has only four triangles.

The first option, i.e. defining a 3D feature as a list of polygons can berealized by two columns in one relational table, i.e. feature_ID and a col­umn for the spatial data type (i.e. polygon). If the reality is quite complex,leading to many 3D features with shaped polygons, the DBMS tableshould be normalized. This means that the polygons have to be organizedin a separate table (containing polygon_ID and a column for the spatialdata type) and the 3D feature table should contain the indices to the com­posing polygons. Clearly, the separate relational table for volumetric ob­jects would be simpler if the volumetric object is tetrahedron. It can be or­ganized in a table with finite number of columns: one for the ID of thetetrahedron and four for the composing polygons (triangles). In general,such an organization can be seen as a partial topological model; since the3D object is defined by references to the composing polygons. Fig. 1 is anexample of a two-table, polyhedron data storage.

In the second approach, a 3D object is stored using the data type multi­polygon, i.e. all the polygons are listed inside the data type, which is prac­tically one record in the relational database. This case requires only one ta­ble, which may contain only two columns: feature_ill and column for thespatial data type. Various examples of these representations can be foundin Stoter and Zlatanova, 2003.

An apparent advantage of the 3D multipolygon approach is the one-to­one correspondence between a record and an object. Furthermore the 3Dmultipolygon (compare to list of polygons) is recognized as one object byfront-end applications (GIS/CAD). For example, a 3D multipolygon isvisualised as grouped polygons in Bentley Microstation. However, in caseof editing, the group still has to be ungrouped into composing polygons(Zlatanova et al 2002), i.e. the group is not recognised as 3D shape.

Indeed, both representations are not recognized by DBMS as a volumet­ric object, i.e. they are still polygons and thus the 3D objects cannot bevalidated. The objects can be indexed as 3D polygons but not as 3D volu­metric objects. Spatial operations can be performed, as well, but on the'flat' polygons, i.e. the z-coordinate will not be considered. Moreover,both representations are highly inefficient in terms of storage space. Thecoordinates of the points in a volumetric object are repeated at least three

3D Geometries in SpatialDBMS 5

times (being part of minimum three neighbouring polygons) either in 3Dmultipolygon record or in the list of polygons.

User defined spatial data types can be implemented using different ap­proaches from the simple SQL create data type statement, to more com­plex implementations, using a Procedural Language (PL) , Java, C++, etc.The common drawback of such an implementation is impossibility to ap­ply the native spat ial functionality (operations and indexing) of DBMS.Moreover the user-defined spatial data types cannot be stored in the samecolumn of the natively supported spatial data types. Visualisation in front­end applications would be possible only by developing individual connec­tions. User-defined spatial data types, however, are very useful for proto­typing for approval of new concepts. Two examples of user-defined spatialdata types are discussed later in the text.

Fig. 2. 3D visualisation of pipelines, organisedas lines in Oracle Spatial

3D line objects

Typical examples of 3D line objects are utility networks: pipeline and ca­ble networks. Utility data and systems have been always predominantlytwo-dimensional. Only recently, investigations have been initiated towardsmaintaining utility networks in three dimensions. Motivation for this is ex­tended usage of underground space and therefore the apparent need of

6 Sisi Zlatanova

more sophisticated mechanisms for visualizing several networks in oneenvironment.

Utility networks (represented as lines with 3D coordinates) can be read­ily managed in DBMS using the supported spatial data types line or multi­line. If required by an application, some point objects (valves, connectors,etc.) can be separately organized as points. The only trouble of 3D lines isthe visualisation in 3D scenes. It is often recommended 3D lines to be sub­stituted with tiny cylinders when rendered. Indeed, such a substitution cannot be justified only for visualisation purposes. Therefore Du and Zla­tanova (2006) suggest keeping the original data as 3D lines and creating3D cylinders on the fly only for the visualization (Fig. 2).

(;lJI,Cl./,0,

~ A,

t~ -.

Fig. 3. 3D visualisation of point clouds , managed as points in DBMS

3D point clouds

Until recently, 3D point objects were relatively rare in real-world data setsand a little attention was given on them. But, with the advances of sensortechnology, laser scanning techniques become available, which producelarge amounts of very specific 3D point data. Theoretically, these pointscan also be organised in DBMS by either 1) using the supported spatialdata types point (Fig. 3) and multipoint or 2) creating a user-defined type.Depending on the type of data (raw or processed data) , the user might de-

3D Geometries in Spatial DBMS 7

cide for either of the representations. Usually the most common format ofprocessed laser scan data consist of seven parameters: x,y,z-coordinates,intensity and colour represented by r,g,b-values. The advantage of pointdata types is the possibility to manage all these attributes for each individ­ual point. The major disadvantage refers data storage and indexing, whichare very expensive (one record per point). The multipoint data type can beefficiently indexed, but the points lose their identification, which might beimportant for modelling purposes (Meijers et al 2005). Furthermore, thenumber of points in one multipoint has to be carefully considered for anacceptable performance (Hoefsloot, 2006). Depending on the point distri­bution and size of the point cloud, the operations can become time con­suming an thus difficult to handle.

New 3D spatial data types

As shown above, 3D real-world feature can be stored and indexed inDBMS and retrieved for visualisation and editing in front-end applicationbut they can be analysed only as 2D features. A true 3D geometry datatype (polyhedron anellor tetrahedron) and corresponding 3D spatial opera­tions (validation, volume, length, intersection, etc.) are missing in allDBMS. Furthermore, the simple 3D volumetric data type would be still in­sufficient for handling many shapes (cone, sphere) available in ABC/CADapplications.

The sections below will briefly present two implementations of new 3Ddata types, i.e. polyhedron and NURBS surface.

3D polyhedron

A 3D spatial data type is the first most important development to be madeby DBMS. A simple 3D object can be represented basically in three differ­ent ways as a polyhedron (consisting of arbitrary number of planar poly­gons which have arbitrary number of points), triangulated polyhedron(consisting of an arbitrary number of triangles) or tetrahedron (composedof four triangles). All the three representations have advantages and disad­vantages (Zlatanova et al 2004). Tetrahedron is the simplest 3D objectconsisting of a finite number of points and triangles. While advantageousfor computations and consistency check (Peninga, 2005), tetrahedrons areless appropriate for modelling of man-made objects such as buildings,bridges, tunnels, etc., because the interior would require a subdivision intotetrahedrons (which should be omitted for visualisation). However, tetra-

8 Sisi Zlatanova

hedrons are widely used in modelling geologic al formations (Breunig andZlatano va, 2006 ). The polyhedron is the most appropriate representationfor man-made objects, but its validation is quite complex (Arens et ai,2005). Triangulated polyhedron is compromise between the two ensuringat least the simplicity of the polygons.

\Fig. 4. 3D polyhedron (Arens 2003)

Arens 2003, and Arens et al 2005, have selected a polyhedron for im­plementation , since it is the most complex data type requiring strict valida­tion rules. A polyhedron is defined as a bounded subset of space, enclosedby a finite set of planar polygons (not self-intersecting) such that everyedge of a polygon is shared by exactly one other polygon. The polyhedronbounds a single volume, i.e. from every point on the boundary, every otherpoint on the boundary can be reached via the interior. The polyhedron hasclearly defined inside/outside space, i.e. it is orientable . The polyhedrondefined in this way can have also cavities (Fig. 4). The polyhedron datatype is implemented in Oracle Spatial object-oriented data model , but theformalism is generic. To avoid duplications of point coordinates (As men­tioned above), the description has two sections: 1) a list of all the point co­ordinates and 2) a sequence of polygons , each composed of a list with in­dices to the point coordinates of the first section. The validity of the newdata type is controlled by a specially designed function, which checks thedefinit ion rules. Several tests were performed on the new data type and theresults were positive.

In very short terms, a tetrahedron data type will be proto typed followingthe formalism and implementation considerations as described in Peningaet a12006.

3D Geometries in Spatial DBMS 9

3D freeform curves and surfaces

Freefonn curves and surfaces such as Bezier, B-spline and NURBS, arebecoming progressively important for modelling of buildings, towers, tun­nels, etc. Very often these models need to be integrated with 3D GIS forinvestigations and adjustment of the construction (e.g. for wind resistance).One option for such an integrated environment can be DBMS. Thereforewe have initiated a research aiming at developing data types that can beused by AEC applications. Among the large amount of mathematical rep­resentations of 3D space (used in CAD and AEC worlds), NURBS is thefirst candidate to be considered. Some of the most important NURBScharacteristics are (Piegl and Tiller, 1997):

• NURBS offer a common mathematical form for both, standardanalytical shapes (e.g. cone, sphere) and free form shapes,

• The shapes described by NURBS can be evaluated reasonablyfast by numerically stable and accurate algorithms,

• Important characteristic for modelling real-world objects is thatthey are invariant under affine as well as perspective transfor­mations.

The only drawback of NURBS is the extra storage needed to define tra­ditional shapes (e.g. circles). Using NURBS data types, a circle can be rep­resented in different ways but the complexity is much higher compared toits mathematical definition (i.e. radius and centre point).

The definition a NURBS curve consists of several quite complex equa­tions, which can be found elsewhere in the literature. Based on these equa­tions, the parameters to be included in the data type are specified as: con­trol points, their weights, knots sequence and a degree value. Theseparameters double in case of surface (Pu, 2005). A NURBS curve requiresfulfilment of a number of conditions such as the degree> 1, the number ofcontrol points >3, Degree = Number of knots - Number of control points ­1, he number of weight values is equal to the number of control points, achweight value> 0, knot vector is non-decreasing and has more than 1 knot.

The new data type has been prototyped for Oracle Spatial, but outsidethe Oracle Spatial SDO_GEOMETRY model, which means it can be read­ily used for any spatial DBMS (PostOIS, MySQL, Informix, etc.). Besidesthe validation function, few simple spatial functions (rotation, translation,scale and distance) were developed. Since the data type is much morecomplex compared to the 3D polyhedron data type, a special care wastaken for the visualisation in ABC software, i.e. an engine was developedto map the NURBS data type to the internal representation of Microstationand AutoCAD (Fig. 5). Two NURBS models were tested with the devel­oped data type for retrieval, editing and posting.

10 Sisi Zlatanova

Fig. 5. NURBS building retrieved from DBMS

Tests have convincingly demonstrated that appropriate data types for ef­ficient management of free form curves and surfaces can be created atDBMS level. The design is compliant with OGC Abstract Specifications(see bellow). But the spatial functionality provided for NURBS data typehas to be very carefully considered. Mat hematics behind many NURBSoperations mig ht appear to complex for implementation at a DB MS level.

Standardization initiatives: Open Geospatial Consortium

With respect to spatial data management and interoperability, several stan ­dardization initiatives have been set off, e.g. ISO, IAI, Web3D, etc . TheOpen Geospatial Consortium will be mentioned here, because its activitieshas largely motivated the spatial developments in RDBMS.

The mission of the OGC, founded in 1994, is to enable interoperabilityof geo-services , OGC produces Abstract Specifications and Implementa­tion Specifications (OGC, 2001). The aim of Abstract Specifications is tocreate and document a conceptual mode l sufficient to create the Implemen­tation Specifications. The Abstract Specifications are subdivided into Top­ics, each of them related to different aspects of geo-spatial services. Topic1 (identical to ISO 19107) is the most important one discussing the spatialschema for representing features. It shou ld be noticed that the Abs tractSpecifications disc uss a wide range primit ives (also those used in CADdomain) such as cone, sphere , triangulated surfaces and freeform shapessuch as B-splines and NURBS. App lying this framework it should be noreal problem representing real world in 3D, using even different represen­tations.

However, the Abstract Specifications provide only the general concep­tual semantics. The way these can be imp lemented at different platforms(based on CORBA, OLE/COM and SQL) is described in three differentImplementation Specifications. The Implementation specifications of in-

3D Geometries in Spatial DBMS 11

terest for DBMS are the Simple Feature Specifications for SQL. Thoseprovide guidance for implementing data types and spatial functions inDBMS. These specifications have greatly contributed to standardizing theimplementing spatial functionality in DBMS. However, they are only 2D,i.e. limited to simple primitives such as points, lines and polygons (sur­faces) and cannot meet the three-dimensional demands.

OGC has already started numerous new initiatives to meet the chal­lenges. Currently 36 projects (working groups) are discussing various as­pects of data integration and exchange only in the Specification Programand almost all of them attempt to handle the third dimension. The mostrelevant is the work of CAD-GIS working group (Case, 2005). The mis­sion of this group is spanning a bridge between CAD, AEC systems andGIS by finding opportunities to improve interoperability of geospatial dataand services across these domains. Incompatibilities at various levels (se­mantic, geometry, topology) contribute to the complexity of the problem(Zlatanova and Prosperi, 2006).

To suggest an appropriate schema for exchange of 3D spatial data, thisgroup will consider several on-going developments, i.e, LandXML (forland survey and construction initiated in 1999 by Autodesk and EAS-Emembers), LandGML, CityGML (GML for city models), aecXML (forAEC including information about projects, documents, materials, parts,organizations, professionals, etc.), TransXML (a project aiming at XMLschemas for exchange of transportation data), IFC (the Industry Founda­tion Classes used to define architectural and construction-related CADgraphic data as 3D real-world objects), OpenFlight (an industry standardreal-time 3D scene description format), 3D ShapeFile (ESRI), X3D, etc.There is a firm believe the work of this group will contribute to extensionof the Implementation Specifications toward real 3D data management.

Concluding remarks

In the last five years DBMS made a large step toward maintenance of ge­ometries as GIS used to manage them. The support of 2D objects with 3Dcoordinates is adopted by all mainstream DBMS. The offered functionsand operations are predominantly in the 2D domain. The DBMS spatialschemas have to be extended to fully represent the third dimension (firstwith simple volumetric object and later with more complex 3D data types).Concepts for 3D objects and prototype implementations are already re­ported, DBMS have to make the next step and natively support them. 3Doperations and functions have to be developed not only for the volumetric

12 Sisi Zlatanova

object but also for all other objects embedded in 3D space. 3D functional­ity is next to be considered. It should be remembered that Spatial DBMS isa place for storage and management, and less intended for extensive analy­ses. The 3D functionality should not be completely taken away from front­end applications such as GIS and CAD/AEC. 3D Spatial DBMS shouldprovide the basic (generic) 3D functions, such as computing volumes andfinding neighbours. Complex analyses have to be attributed to the applica­tions.

Some existing data types are clearly not sufficient for the purpose ofsome applications. A very typical example is the multipoint. It was defi­nitely not designed for large amounts of points as from laser scanning.DBMS fail to handle efficiently such amounts of data until now. Suchpoints need a special treatment. On the one hand, with the progress of datacollection techniques, the amounts of points will only increase. Many la­ser-scanning companies are increasingly getting concerned about the man­agement of such data. On the other hand, the advances in 3D modellingwould require more intelligent management of both row and processeddata. Clearly a new spatial data type with internal structure and index hasto be developed.

Triangle (or TIN) data type is also quite demanded. It is likely that TINwill continue to be widely used for all kinds of complex surface represen­tations in GIS. Most of the terrain representations presently maintained inGIS as well as many CAD designs (meshes) are TIN representations. TINscan be stored in DBMS using the polygon data type. This data type is gen­erally assumed for multiple vertices and thus over-attributed. Thus a sim­pler adapted data type is required.

Maintenance of multiple representations to be used as Level-Of-Detailsis another critical aspect of large three-dimensional models. Still the man­agement of multiple representations is far from formalized. A very promis­ing initiative is CityGMI.." which concepts can further be incorporated inthe Spatial Schema and later incorporated in Implementation Specifica­tions.

Management of texture and mechanism for texture mapping and texturedraping is critical for management of realistic 3D City models. 3D objectsusually need more attributes for visualisation compared to 2D objects.Moreover, very often 3D objects are textured with images from real world.As AEC and GIS applications come together the question of linking tex­tures to geometries will appeal. Textures can be understood as 'presenta­tion' attributes of 3D objects an also decoded in the data types.

As mentioned above, the update of the OGC Implementations Specifica­tions is very critical. Standardized vision on 3D objects and functions onthem will stimulate DBMS toward 3D implementations. The first exten-