Geo-spatial Information Science

ISSN: 1009-5020 (Print) 1993-5153 (Online) Journal homepage: http://www.tandfonline.com/loi/tgsi20

Design and implementation of 3D model database

for general-purpose 3D GIS

Weiping Xu , Qing Zhu , Zhiqiang Du & Yeting Zhang

To cite this article: Weiping Xu , Qing Zhu , Zhiqiang Du & Yeting Zhang (2010) Design and
implementation of 3D model database for general-purpose 3D GIS, Geo-spatial Information
Science, 13:3, 210-215, DOI: 10.1007/s11806-010-0309-7

To link to this article: https://doi.org/10.1007/s11806-010-0309-7

Copyright Taylor and Francis Group, LLC

Published online: 14 Aug 2012.

Submit your article to this journal

Article views: 639

Citing articles: 7 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=tgsi20
Geo-spatial Information Science 13(3):210-215 Volume 13, Issue 3
DOI 10.1007/s11806-010-0309-7 September 2010

Article ID: 1009-5020(2010)03-210-06 Document code: A

Design and Implementation of 3D Model Database

for General-Purpose 3D GIS

XU Weiping, ZHU Qing, DU Zhiqiang, ZHANG Yeting

State Key Laboratory of Information Engineering in Surveying Mapping and Remote Sensing, Wuhan University,129 Luoyu
Road, Wuhan 430079, China

© Wuhan University and Springer-Verlag Berlin Heidelberg 2010

Abstract To improve the reusability of three-dimensional (3D) models and simplify the complexity of natural scene reconstruc-
tion, this paper presents a 3D model database for universal 3D GIS. After the introduction of its extensible function architecture,
accompanied by the conclusion of implicit spatial-temporal hierarchy of models in any reconstructed scene of 3D GIS for general
purpose, several key issues are discussed in detail, such as the storage and management of 3D models and related retrieval and load
method, as well as the interfaces for further on-demand development. Finally, the validity and feasibility of this model database are
proved through its application in the development of 3D visualization system of railway operation.
Keywords 3D model database; 3D GIS; 3D modeling; CAD models
CLC number P208

versal 3D GIS lies in the improvement on the reus-

Introduction ability of existing models like 3D CAD models, just
as 2D symbolization plays an important role in the
The accurate and efficient reconstruction of 3D vir- enhancement of information communication and ac-
tual geographic environments has always been a focus celeration of mapping.[4] Especially, there is an in-
of research in 3D GIS field. In general, most methods creasing trend of CAD’s combination with 3D GIS
for rapid modeling can be divided into two categories: because of its powerful modeling and editing capa-
one is the use of existing geospatial techniques like bilities. For example, CAD models of individual
LIDAR and digital photogrammetry to quickly acquire buildings can be used to make up the shortage of
original 3D geospatial data, which is subsequently measured survey and terrestrial photogrammetry,
processed in a lot of complex editing work, while the which may have difficulty in obtaining complete
other is based on interactive modeling from reusable building information due to environmental restraints
computer-aided design (CAD) models, which may be like occlusion between objects.[5] Besides, CAD
extracted from constructed projects or imported from models can represent more complex structure, finer
CAD modeling software.[1-3] material features, and higher level of details of enti-
Nowadays, the key to the promotion of the low ef- ties. Several popular formats of them and related in-
ficiency of modeling in scene reconstruction for uni- troduction are listed in Table 1.

► Received on February 20, 2010.

► Supported by the National Natural Science Foundation of China (No.40871212, No.40671158); the National High Technology Research and Devel-
opment Program of China (No.2008AA121600).
► XU Weiping now is studying for doctor degree in photogrammetry and remote sensing in Wuhan University. His current research interests include
modeling methodologies for virtual geographic environment and application of 3D GIS.
► E-mail: [email protected]
 XU Weiping, et al./ Design and Implementation of 3D Model … 211

Table 1 Introduction of five popular formats of CAD models

File format Software / Program Creator Graphic information contained Generality
DXF AutoCAD AutoDesk Geometry; Color Good
Geometry; Material; Texture; Lighting;
3DS 3DSMAX AutoDesk Much better
Animation; Camera
Geometry; Material; Texture; Lighting;
FLT Multigen Multigen Much better
Animation; Camera
OBJ Wavefront Wavefront Geometry; Material; Texture Common
WRL VRML Silicon Graphics Geometry; Material; Texture; Lighting Better

In practice, simple block models, derived from 2D model in pursuit of high level of details is often fairly
map data in regular modeling, such as building extru- large, while GIS needs to deal with a lot of such enti-
sion without roof, are available from graphic design- ties at the same time. Then, it is considered desirable
ing and planning softwares like AutoCAD. Another to simplify the original model or reorganize its stor-
type of model, mesh model, used widely in 3D scene age structure of 3D graphic information for the bal-
visualization and spatial analysis, generally contains ance among rendering efficiency, data scheduling,
detailed geometrical and simple topological informa- and graphic quality in 3D GIS.[2,8] (3) Most CAD
tion.[6,7] Fig.1 shows a 3D mesh model in a 3DS for- models are independent in 3D GIS and cannot be
mat of a Chinese building. used to present an object’s temporal behavior on
which 4D GIS is given more and more attention to.
Aiming at the design of 3D model database for
general-purpose 3D GIS and the solution of critical
issues mentioned above, this paper presents its ability
to import/export existing CAD models in popular
formats, store similar models in form of specific
Fig.1 3D solid and wireframe modes of a building
groups, and thereby reuse them in different applica-
One proposal of integration of CAD to 3D GIS is tion that reduces a great deal of time investment in
to establish 3D model database based on library ap- modeling. Section 1 gives the function architecture of
proach to meet the changing needs of model reuse for this database and is followed by details of the storage
general-purpose 3D GIS.[2] Moreover, it is wise to and management of database in Section 2. The
enrich and enlarge the database in batches of model- method of model retrieval and load is discussed in
ing, recommended as the strategy of model collection. Section 3. At last, its implementation and application
In this approach, the idea of reusability suggests the in railway visualization is illustrated.
possibility for 3D models, especially complex and
detailed models, to be completely extracted from re- 1 Function architecture of 3D model
constructed scenes and stored in an open data base,
database
which other modelers can be allowed quick and easy
access to for their own modeling need. This is greatly
1.1 Data hierarchy of models
desirable and feasible, the moment Internet has per-
meated many fields of social development, where a 3D models can be graphic or image models for the
number of various models are available now for free description of shapes, positions, orientations, and
or for a fee.[9] sizes of geographical entities, as well as the expres-
In this integration, there are several critical issues sion of aspects of their spatial-temporal distribution
that needs to be considered: (1) Modelers have to link and dynamic changes.[4] It can be concluded in the
imported models with given attributes in GIS, for any following respects that most of them included in a
dataset of GIS is composed of spatial data associated complicated scene often form an implicit spatial and
with nonspatial properties that original CAD models temporal hierarchy based on fundamental models.
are lacking in. (2) The data amount of single CAD z The distribution of independent models in accor-
212 Geo-spatial Information Science 13(3):210-215

dance with fixed rules in one piece of scene. Most of objects’ behavior fall into two categories: definite
these rules can be parameterized. For example, a po- behavior with known states in discrete time like
lygonal vegetation area may be covered by several turnout’s switch and uncertain behavior in response to
types of trees and bushes of different sizes, distributed real time events, such as collisions.[10] CAD models
in regular rows and columns, or random positions with are generally stationary and thus applied more in
defined density to lighten inaccurate modeling work- former presentation. In a broad sense, the transition of
load (Fig.2). levels of detail (LOD) can also be regarded as a type
of behavior just influenced by human’s sight but not
real world.

1.2 Function architecture

Considering the hierarchy stated above, the paper

presents the function architecture of 3D model data-
base for general-purpose 3D GIS (Fig.5). It mainly
includes three parts: import and export, interactive
Fig.2 A polygonal vegetation area covered by two visualization, and database management to be dis-
types of trees cussed in Section 2.
z The composition of submodels or component
models in one whole model. Not only these sub-
models can just be grouped with different spatial po-
sitions, scales, and directions to compose an “assem-
bly model” or a geometrically regular model like
lofting, but also one gray model can be bind with
corresponding texture submodels to enhance the real-
ism of landscape (Figs. 3 and 4).

Fig.5 Function architecture

z Format conversion. As stated previously, it is nec-

essary to simplify the model or reorganize 3D graphic
Fig.3 A landscape model of signal lamp composed
of gray model and bind textures information contained in CAD model. However, com-
plete automated simplification is time-consuming.
One feasible way is to store vertex’s properties in tri-
angle-strip structure stored in user-defined format,
which can greatly improve rendering efficiency in 3D
GIS.
z Data encryption/decryption. To protect confi-
dential model data from theft, users can choose to en-
crypt the model file and decrypt it in the proper time.
Fig.4 A pavement created through loft of roadbed z Batch import/export. It relieves greatly the sub-
section along road centerline
mission to database of a batch of models, such as a
z The combination of state models to present an large file directory of classifiable model files by their
object’s temporal behavior. In general, the types of names, to establish automatically a tree structure of
 XU Weiping, et al./ Design and Implementation of 3D Model … 213

classification. On the contrary, users can also export a tribute table by MID and its linked SGID, which can
specific class of models ranging from formats or fea- be used to acquire other state models in state table.
tures to share with others.
Table 2 Detailed structure of attribute table
z Perspective view. It is for browsing of model’s
Field name Field type Field description
contents stored in converted files in solid or wire-
MID INTEGER Model’s ID
frame mode.
CID INTEGER ID of class the model belongs to
z Thumbnail preview. A model’s thumbnail in fa-
MNAME STRING Model’s name
vor of later scanning is created at the time of its initial FORMAT INTEGER Format of model file
visualization, with its image data stored in model ATTRIS STRING Linked attribute
format or standard image format like jpg/jpeg. Be- ID of state group the model is
SGID INTEGER
sides, users can choose a proper vision angle in per- Packed in
spective view for thumbnail re-creation.
Table 3 Detailed structure of class table
z Animation /exhibition. Smooth transitions of
Field name Field type Field description
selected models for exhibition can be connected by
CID INTEGER Class ID
simple rotation or other special animation effects
CNAME INTEGER Class name
like bounce. This is the same as a series of 2D pic-
SUBNUM INTEGER Number of included subclasses
tures made into sliders smoothed by shutter or mosaic
SUBIDS BLOB Included subclasses’ ID datasets
effect.
Table 4 Detailed structure of state table
2 Database design Field name Field type Field Description
SGID INTEGER ID of state group
Considering the function architecture of this model SGNAME STRING Name of state group
database presented in Section 1, its storage structure SDESC STRING State description
MIDS BLOB packed models’ ID dataset
can be divided into two parts: model files and infor-
mation tables (Fig.6). The former may be a group of
one whole directory and its subdirectories of im- 3 Model retrieval and load method
ported models in original or converted files. The
models’ given properties from manual attribute re- Currently, most of 3D model databases provide the
cording are stored in the attribute table, while their function of text-based model search for key words in
class and state information can be obtained from the filenames, formats, attributes and so on, which cannot
class table and state table, recorded when each model’s work in some cases, such as when objects are not an-
category and state models’ package is customized. notated or annotated with unspecified or derivative
These tables can be efficiently established and que- keywords.[11] Therefore, the content-based multime-
ried in commercial database like SQL Server. dia retrieval method is emerging for models’ shapes
analysis as well as automatic classification and iden-
tification.[12] The 3D model search engines adopting
this technique allow the input of 2D sketches or pho-
tos instead of text (Fig.7), which accord with user’s
direct impression of objects.[13,14] One of its important
procedures is to establish 2D and 3D geometrical
Fig.6 Storage structure of database
shape feature index from feature extraction and match
The detailed information of three tables is listed as them according to similarity measurement. However,
follows (Tables 2, 3, and 4). They are linked by re- it is not appropriate in practice for the efficiency of its
lated fields. For example, if one model’s MID is search for a batch of models instead of a single model
known and other models in the same state package cannot meet the needs of retrieval and load in large-
are wanted, users can fast retrieve the model from at- scale scene reconstruction.
214 Geo-spatial Information Science 13(3):210-215

4 The implementation and application

This research was undertaken in a 1024 Mb RAM

2.26 GHz Inter Pentium PC with an ATI MOBILITY
FireGL 3200 graphics card. The operation system is
Windows XP, and three information tables mentioned
in Section 2 are stored in Microsoft Office Access 2003.
The import and export interface (input and output in-
terfaces) supports the conversion from formats of 3DS,
DXF, and WRL to SUP, which adopts strip storage of
Fig.7 2D sketch query interface of one 3D model search
engine vertices. Its file structure is listed in Fig.9.
File header
In pursue of fast and automatic retrieval to make ID of Format Scope of all models’
Origin coordinates
version boxes
model load easy, the further-development interfaces
File body
of the 3D model database are given in the paper, in- Num of models Scope of each model’s box
Num of vertices List of vertices’ coordinates
cluding export/import interface, fuzzy query-by-text
Num of normal vectors List of normal vectors
interface, and interactive selection interface. For ex- Num of texture coordinates List of texture coordinates
ample, developers are suggested first to design their Num of material List of each material’s data
Num of lights List of each light’s parameters
own text files for configuration of models’ geometri-
Fig.9 File structure of SUP format
cal and attribute information and then submit them to
the database, waiting for the corresponding result set The main UI of this 3D model database is designed
of models from automatic search by key words consistent with resource manager in Windows, which
through query interface. Afterward, the pattern for can be easily familiarized by the majority of users
interactive selection of satisfactory models from the (Fig.10). It is divided into a tree view for classifica-
result set can also be customized via selection inter- tion, a list view for thumbnails, two perspective
face. Next, all selected models need to be loaded to views in solid and wireframe mode, and an attribute
the scene from the database by export/import inter- editing view.
face and grouped in different ways described in Sec-
tion 1. Suppose the real-time simulation of water pipe.
Followed by each valve’s correct placement and ori-
entation in stationary modeling, its on-off state model
package can be retrieved to guide its open/ close be-
havior. The whole application process stated above is
expressed in Fig.8.

Fig.10 Main UI

In the development of 3D visualization system of

railway operation, the model database was success-
fully applied in modeling work. In general, the mod-
els contained in railway scene can be classified into
several types as follows:
Fig.8 Process of further-development application Lofting models. In general, loft function only re-
 XU Weiping, et al./ Design and Implementation of 3D Model … 215

quires a section surface to scan along a path to con- the railway station can be regarded as a composition
struct a model. Here, sections of railway bed (includ- of two poles and many units distributed along the line
ing rail), bridge body, and tunnel are quickly retrieved between the poles (Fig.11(c)). In a similar way, a
from the database through query interface (Fig.11(a)). bridge is held up by a number of piers at fixed inter-
Independent models. Turnout, intersecting cross- vals parallel with the bridge body.
over, and signal lamp fall into this category (Fig.11(b)).
Their positions are all obtained from the configura- 5 Conclusion
tion file, while the orientations need to be computed.
For example, the signal lamp’s orientation is equated The main research contributions are (1) extensible
to the closest railway for their parallelism. Besides, function architecture in support of spatial-temporal hi-
their behavior needs to be considered. Compared with erarchy of models and (2) fast and automatic load
two kinds of a turnout’s state, an intersecting cross- method from the database to the scene with the help
over contains eight as each of its four nodes may be of further-development interfaces. Such kind of 3D
open or closed. model database would become more and more im-
Parameterized models. The catenary system near portant in the modeling of 3D GIS.

Fig.11 Models in the railway scene

Aided Design, 29(4): 321-330

References
[8] Luebke D, Reddy M, Cohen J D, et al. (2003) Level of
[1] Fujii K, Arikawa T (2002) Urban object reconstruction detail for 3D graphics[M]. San Francisco: Morgan
using airborne laser elevation image and aerial image[J]. Kaufmann
Geoscience and Remote Sensing, 40(10): 2234-2240 [9] Costello P J, Bee S T (1997) 3D model databases: The
[2] Whyte J, Bouchlaghem N, Thorpe A, et al. (2000) From availability of 3D models on the World Wide Web [OL].
CAD to virtual reality: modelling approaches, data exchange http://www.agocg.ac.uk/reports/ virtual/36/report36.html
and interactive 3D building design tools[J]. Automation in [10] Faisstnauer C (1998) Navigation and intersection in virtual
Construction, 10: 43-55 environments[OL]. http://www.cg.tuwien.ac/at/~faisst/pub/
[3] Bosche F, Haas C T (2008) Automated retrieval of 3D CAD MasterThesis
model objects in construction range images[J]. Automation [11] Funkhouser T, Min P, Kazhdan M, et al. (2003) A search
in Construction, 17(4): 499-512 engine for 3D models[J]. ACM Transactions on Graphics,
[4] Zhu Q, Gao Y, Wei Y, et al. (2003) Design of 3D models 22(1): 83-105
for 3D GIS[J]. Geomatics and Information Science of [12] Yang Y, Lin H, Zhu Q (2004) Content-based 3D model
Wuhan University, 28(3): 283-287 retrieval: a survey[J]. Chinese Journal of Computers, 27(10):
[5] Gong J, Zhu Q, Sui H, et al. (2003) Key issues for data 1297-1310
transformation from CAD to GIS[J]. Engineering Journal of [13] Min P (2004) A 3D model search engine[D]. Princeton:
Wuhan University, 36(3): 64-68 Princeton University
[6] Weihe K, Willhalm T (1997) Reconstructing the topology of [14] Ansary T F, Vandeborre J P, Daoudi M (2007) 3D-model
a CAD model-a discrete approach[R]. Lecture Notes in search engine from photos[C]. Proceedings of the 6th ACM
Computer Science, 1284 International Conference on Image and Video Retrieval,
[7] Fitzgibbon A W, Eggert D, Fisher R B (1997) High-level Amsterdam
CAD model acquisition from range images[J]. Computer-