IJCSNS International Journal of Computer Science and Network Security, VOL.8 No.

1, January 2008 49

Research and Development of 3D Modeling

Xi-Dao LUAN†, Yu-Xiang XIE, Long YING and Ling-Da WU
†
School of Information System and Management, National University of Defense Technology, Changsha 410073,
China
Summary years, introduces applications of 3D modeling including
3D modeling is a key technique to much research and applications. tissue engineering and heritage protection. Finally, several
In different fields of research and applications, 3D modeling main problems and a few deficiencies are pointed out and
methods of model-data acquisition and modeling have their own further challenges foreseen from three aspects: modeling
specialties. The paper systemically introduces equipments of 3D
retrieval, digitizing method, and dynamic modeling.
data acquisition and modeling methods, discusses the characters
and developments of laser scanning system and Image-Based
Modeling and Rendering (IBMR) in recent years. The paper also
introduces applications of 3D modeling including tissue
2. 3D data acquisition
engineering and heritage protection. Finally, several main problems
and a few deficiencies are pointed out and further challenges To model an object, it is necessary to get data about
foreseen from three aspects: modeling retrieval, digitizing method, object’s size and depth at first. In different fields of research
and dynamic modeling. and applications, there are different demands accordingly.
Key words: Equipments and modeling methods adopted in the
3D modeling; IBMR; Data Acquisition; Rendering modeling process are also different. For example, to
reconstruct 3D underwater environment, people have to
make use of sonar and ocean satellite to get the underwater
1. Introduction terrain data. To improve the understanding of a human
operator driving an underwater Remotely Operated Vehicle,
Building exact models to simulate and present principles of U.Castellani reconstructs 3D underwater environment by
things’ behaving, is a general and important method for acoustic camera system in real time, and the resolution is
modern science. In applications and science area, there is about 5cm at the frequency of the acoustic signal
also a growing interest and trend in modeling the world in 500kHz[1].
digital three dimensions. To model and simulate the real In general, measurement systems comprise of contact
world, it is a process of digitizing objects' shape, motion, and non-contact methods [5]. The contact methods vary
texture and other properties according to the emphasis. from simple measurement using tape to the sophisticated
Much work has been done and has received Coordinate Measuring Machine (CMM). These two
remarkable achievement on digital equipments and methods are commonly used in mechanical engineering.
software. Computer-aided 3D modeling tools, including Measurement using tape is conventional and subject to
laser scanning system and Image-Based Modeling and errors. Its efficiency is slow. Though CMM is totally
Rendering (IBMR), are getting more and more powerful. automated and its measurement accuracy is 0.02mm, CMM
Modeling objects evolve from objects with simple structure requires a stable platform and the object’s size is limited.
to complicated human face, limb, hair, and even fluid. Technology of non-contact measurement is a result of
In the past, the main applications of 3D modeling the development of computer graphic and vision, acoustics,
were visual inspection and robot guidance. Nowadays, the optics and related equipments. Non-contact methods
emphasis is shifting. 3D modeling has been widely used in include laser scanning system (e.g. VIVID910 laser
areas including computer graphics, virtual reality and scanner), geodetic total station system (e.g. AXYZ system,
communication, and there is more and more demand for 3D the accuracy is about 0.05mm), close range
content for tissue engineering and heritage protection. In photogrammetric system (e.g. V-STARS system, its
different fields of research and applications, 3D modeling accuracy is better than 0.05mm), and structured light
methods of model-data acquisition and modeling have their system (e.g. Eyetronics’ ShapeCam[2], whose accuracy is
own specialties. up to 0.5mm based on how tight user focuses his grid),etc
In this paper, the process of 3D modeling including [5].
3D data acquisition, modeling and rendering is introduced
systemically. The paper also discusses the characters and
developments of laser scanning system and IBMR in recent

Manuscript received January 5, 2008

Manuscript revised January 20, 2008
50 IJCSNS International Journal of Computer Science and Network Security, VOL.8 No.1, January 2008

2.1 Laser Scanning System Dhenain et al. scann mouse embryos by micro-MRI, and
the resolution achieved is 20-80 micron voxels.
Laser scanning systems, for example VIVID910, have It is easier for IBMR to map a suitable texture on
remarkable advantages: high speed (scanning time varies model’s surface. But effect of this method mainly depends
from 0.3 sec (fast mode) to 2.5 sec (fine mode)), enormous on the algorithmic design of modeling. These algorithms,
data (up to 77,000 or 300,000 points for fast and fine such as modeling based on shadow or silhouettes, are
modes respectively), high accuracy (0.22mm, 0.16mm and mostly complex and need more computing resource. In
0.10mm in x, y and z respectively), and simplicity. addition, this type of algorithm can hardly model accurately
According to object sizes and measurement distances, users an object with a concave part of surface. It is a common
also can select and use one of 3 types of exchangeable issue for perspective-based algorithms.
mounted lens. In recent years, to acquire 3D information in
To model outdoors environments efficiently and laboratory, some researchers adopt common equipments
accurately, Toshihiro and Masayuki [3] get range and color including a turntable and a digital camera [8, 9, 10]. The
data by integrating an omnidirectional laser rangefinder and components are not complex and easy to find, install, and
an omnidirectional multi-camera system. To register use. In these systems, the cameras are all fixed. By turning
multiple range data of wide area stably and simultaneously the turntable, obtain the real object’s multiview range
by the improved ICP algorithm, planar surfaces such as images. To register and merge range images into a common
walls and roads are extracted and used in the registration coordinate system automatically, Soon-Yong Park et al.
process. The generated surface model is then texture- introduce and calibrate a turntable coordinate system with
mapped by omnidirectional images selected in respect to the camera coordinate system [8]. Ulaş Yılmaz et
consideration of resolution and occlusion. The al. propose a vision based camera calibration algorithm and
omnidirectional laser rangefinder LMS-Z360’s measurable use this algorithm in the extraction of the rotations axis [9].
range varies 1m to 200m, and its measurement accuracy is Miguel Sainz et al. [10] and [8] add artificial fiducial
±12mm. patterns to the scene to help detect features and calibrate the
To acquire 3D information of buildings in an accurate stereo camera. The hardware configuration is simple and
and fast way, Yusuf Arayici adopted 3D laser scanner inexpensive. The complexity of the methods is also
technology for CAD modeling [4]. In this EU funded decreased. But the design of these systems limits their
INTELCITIES project, the CAD modeling is integrated application, especially for wide area environment and
with various systems such as 3D printing and VR objects that can’t be moved.
projection systems for storing data of existing buildings, e- Before constructing models, calibrating camera to get
Planning and e-Inclusion, e-Regeneration, Virtual Urban camera’s internal and external parameter is important and
Planning, etc (www.intelcitiesproject.com). necessary. For systems that use fixed cameras, calibrating is
Tough this type system can get 3D data of objects’ easier, but some parts of object with key geometry
surface as point clouds in an accurate, fast and automatic characters may not be sampled and acquired. The model
way, the related equipments are expensive and restricted to reconstructed will have inevitable deficiency of structure
the condition; some systems do not work well in sun or rain. and texture. Compared with fixed-camera system, systems
Though the acquisition phase can profit of automatic that use hand-held cameras can get object’s 3D data of all
procedures, the 3D point clouds needs post-processing for sides, but it brings difficulties of camera calibration and
a useable output. To represent an architectural object image matching.
analytically, it especially requires many manual actions. To get hand-held camera’s parameter, Liu Gang [11]
accordingly selected some characteristic points of range
2.2 Close Range Photogrammetric System images by human-computer interaction. This type of system
also can input some patterns in the scene or the background
Close range photogrammetric systems, such as V-STARS to help calibration, while the object to be modeled might
system, are relatively cheap. Research results of [5] indicate occlude partially the pattern. Anselmo Antunes Montenegro
its practicality in terms of speed, reliability, flexibility and et al. [12] adopt a calibration method based on model
accuracy. In medical area, 3D data of organs and tissue recognition, which is an adaptive space-carving algorithm
usually come from computed tomography (CT), magnetic that uses photometric and segmentation information.
resonance imaging (MRI), optical microscopy, micro CT,
etc. Each has its own advantages and limitations. The latest
development of micro-CT technology has been successfully
used to model lung tissue at 10-50 micron resolution [6].
IJCSNS International Journal of Computer Science and Network Security, VOL.8 No.1, January 2008 51

3. 3D modeling and Rendering Range Images

Graphics-based modeling and rendering (GBMR) is a

traditional method. 3D modeling software, such as Feature Extraction Camera Calibration
and Matching
AutoCAD and Maya, can help modeling by interaction. But
it is time-consuming and demands many skills for users,
especially to model scenes with complicated and irregular Projection
Reconstruction
Visual
Hull
Space
Carving
... Image-Based Modeling

structures.
Surface
3.1 Image-Based Modeling and Rendering Reconstruction

Viewport Plenoptic Light Field View

While traditionally computer graphics focuses on Interpolation Function Rendering Morphing ... Image-Based Rendering

transforming 3D data into 2D image projections, images

play a more central role in IBMR. This field is
comparatively young; its configuration is simple and mostly 3D model

uses common equipments; it can model objects with

different sizes; it also can get the information of depth and Fig. 1 Image-Based Modeling and Rendering (IBMR)
texture at the same time. IBMR is still a focus in 3D
modeling field. Next step of modeling includes different methods (Fig.
Carlos Hernández Esteban [13] classifies methods of 1), such as projection reconstruction, visual hull, space
IBMR according to the information they use. The first class carving. [15] proposes an algebraic dual-space method that
is methods of shaping from silhouette. These methods requires no correspondences. The method estimates depth
obtain an initial estimation of 3D model known as visual information on contours from curvature information and
hull. They are robust and fast, but because of the type of creates a visual hull of the object. The method also requires
information these methods use, they are limited to simple- that surfaces should be sampled more densely where
shaped objects. The second class includes methods of curvature changes rapidly.
shaping from shading. These methods are based on the In traditional, model’s surface is constructed by
diffusing properties of Lambertian surfaces, and very polygon mesh. While in many 3D information acquisition
dependent on the light condition. The third class is the systems, the 3D data got is point data or point clouds. As
methods that use the color information of the scene. There point clouds do not possess topological connectivity
exists different ways of using the color depending on the explicitly, they are proven to be more convenient to
type of the scene to be reconstructed. One way is to carve a simplify latter processing and presenting. They are more
voxel volume by color consistency, while the result is a suitable in applications that do not require continuous
model composed of a set of voxels, which is difficult to surface or perform view-dependent multi-resolution re-
represent a 3D mesh model. Another way is to guide a sampling.
deformable model by the color consistency, which is IBMR can map texture on object’s surface easily to
sensitive to light condition for these methods’ comparing synthesize a realistic model [10, 11, 16]. During the
absolute color values. The other way is to compare local process of image-based rendering (IBR), operations need
variations of the texture, such as cross-correlation methods not much computing resource; the rendering time is
described in [14]. Some methods of using color independent of the complexity of the scene, and the method
information use some other type of information at the same can synthesize realistic scenes and animation in real time.
time, such as texture and silhouette integrated in [13]. Most methods of IBMR are independent of scale, so it can
Though these methods can achieve a better model, the obtain a 3D model of the whole site or object. IBR mainly
quality is still limited for the way of merging different data. include methods of viewport interpolation, plenoptic
function, view morphing, light field rendering, concentric
3.2 Researches and Development of IBMR mosaics.
When modeling outdoor environment and buildings,
After getting range images, the first step is calibrating the [3] and [17] integrate laser scanning and IBMR these two
camera by feature extraction and matching. Normally, in different types of method, which can not only get point
this step, select and match at least 7 pairs of corresponding clouds in an automatic, fast and accurate way, but also
points between two images, calculate the fundamental present models in detail, which is the advantage of IBMR in
matrix by polar geometry constrain, and get projection texture mapping. So, these two methods are complementary
matrix at last. in a certain sense.
52 IJCSNS International Journal of Computer Science and Network Security, VOL.8 No.1, January 2008

4. Applications of 3D modeling References

[1] U. Castellani, A. Fusiello, V. Murino, L. Papaleo, E. Puppo,
Nowadays, 3D Models are used in a large range of exciting M. Pittore. A complete system for on-line 3D modelling from
applications areas: Animation, Archaeology, Architecture, acoustic images. Signal Processing: Image Communication,
Dentistry, Education, Fashion and Textiles, Foot Wear, 2005, 20: 832–852.
Forensics, Games, Industrial Design, Manufacturing, [2] Eyetronics, 2004, http://www.eyetronics.com
[3] Toshihiro ASAI, Masayuki KANBARA, Naokazu
Medical, Movies, Multimedia, Museums, As-built Plants
YOKOYA. 3D Modeling of Outdoor Environments by
Rapid Prototyping, Reverse Engineering, Sculpture, Toys, Integrating Omnidirectional Range and Color Images.
Mold Making, and Web Design. 3D modeling has become Proceedings of the Fifth International Conference on 3-D
a key technology in many applications. Digital Imaging and Modeling (3DIM’05)
Research on 3D modeling promotes the development [4] Yusuf Arayici, Andy Hamilton. Modeling 3D Scanned Data
of heritage protection. There exist some projects of heritage to Visualize the Built Environment. Proceedings of the Ninth
protection that use laser scanning, such as Digital International Conference on Information Visualisation, 2005,
Michelangelo Project [18, 19], Stanford Digital Formae 509 ～514
Urbis Romae Project [20]. [21] uses ShapeCam, a [5] Halim SETAN, Mohd Sharuddin IBRAHIM. Close Range
Measurement and 3D Modeling. Presented at the 1st
structured light system, to build a detailed 3D model of the
International Symposium on Engineering Surveys for
Antonineny mphaeum at the ancient city of Sagalassos Construction Works and Structural Engineering, 2004
(SW-Turkey). Pagpagiannakis et al. built Turkey Istanbul’s [6] Kriete A., Breithecker A., Rau W. 3D imaging of lung tissue
internal and external models of SS. Sergius and Bachhus by confocal microscopy and micro-CT. Proceedings of SPIE -
Church by 3D Studio Max. The International Society for Optical Engineering, 2001,
Computer-Aided Tissue Engineering (CATE) is a 469 ～476
new field of biomedicine. In this field, 3D modeling is [7] Dhenain M, Ruffins SW, Jacobs RE. Three-dimensional
widely used to help modeling tissue and organs, guide the digital mouse atlas using high resolution MRI.
design of artificial organs and tissue, such as bones and Developmental Biology, 2001, 232(2): 458 ～470
[8] Soon-Yong Park, Murali Subbarao. A multiview 3D
vessel, and perform tissue engineering [22, 23, 24].
modeling system based on stereo vision techniques. Machine
Vision and Applications, 2005, 16: 148 ～156
[9] Ulaş Yılmaz, Adem Mülayim, Volkan Atalay. Reconstruction
5.Conclusion and discussion of Three Dimensional Models from Real Images. Proceedings
of the First International Symposium on 3D Data Processing
More and more fields now need and adopt technologies of Visualization and Transmission (3DPVT.02),2002
3D modeling. There are a number of directions in which [10] Miguel Sainz, Renato Pajarola, Albert Mercade. A Simple
we need to continue. Foremost among these is 3D model Approach for Point-Based Object Capturing and Rendering.
retrieval. To measure similarity of models, characters of IEEE Computer Graphics and Applications, 2004,
3D models, such as shape, topological construction and July/August: 24 ～33
texture, are used. These characters are difficult to describe [11] Liu Gang, Wang Zhangye, Peng Quensheng. Generating
for users and complicated to calculate, while an effective Visual Hulls From Freely Moving Camera. Journal of
Computer-Aided Design & Computer Graphics,
retrieval function is necessary for an integrated 3D
2004,16(11),1501 ～1505 (in Chinese)
modeling system. [12] Anselmo Antunes Montenegro, Paulo C.P. Carvalho, Luiz
Nowadays, the visual quality becomes one of the main Velho, Marcelo Gattass. Space carving with a hand-held
points of attention. There is more and more demand for 3D camera. Proceedings of the XVII Brazilian Symposium on
content with higher accuracy. Information of scene and Computer Graphics and Image Processing (SIBGRAPI’04),
object could not be collected absolutely during the 3D data 2004, 396～403
acquisition, and some data is inevitably lost, we could not [13] Carlos Hernández Esteban, Francis Schmitt. Silhouette and
recover the real word from videos or images by the current stereo fusion for 3D object modeling. Computer Vision and
design. So, it is worthy for us to explore new methods to Image Understanding, 2004, 96: 367～392
[14] A. Sarti, S. Tubaro. Image based multiresolution implicit
digitize the real world.
object modeling. EURASIP J. Appl. Signal Process, 2002,
Dynamic model is our new direction for the future 10: 1053～1066
work. Dynamic models can simulate reciprocal actions of [15] Matthew Brand, Kongbin Kang, David B. Cooper. Algebraic
objects, which is also very helpful in exploring the solution for the visual hull. Proceedings of the 2004 IEEE
discipline of thing’s evolvement. Computer Society Conference on Computer Vision and
Pattern Recognition, CVPR 2004, I33 ～I35
[16] Yoo-Kil Yang, Jung Lee, Soo-Kyun Kim, Chang-Hun Kim.
Adaptive Space Carving with Texture Mapping. LNCS 3482,
2005, 1129～1138
IJCSNS International Journal of Computer Science and Network Security, VOL.8 No.1, January 2008 53

[17] Livio De Luca, Philippe Veron, Michel Florenzano. Reverse

engineering of architectural buildings based on a hybrid LingDa WU born in 1962. Ph.D,
modeling approach. Computers & Graphics, 2006, 30: professor and doctor supervisor in
160～176 Multimedia R&D Center, National
[18] Digital Michelangelo project, University of Defense Technology. Her
http://graphics.stanford.edu/data/mich/ main research interests are virtual reality
[19] M. Levoy, K. Pulli, B. Curless, S. Rusinkiewicz, D. Koller, L. and multimedia technology.
Pereira, M. Ginzton, S. Anderson, J. Davis, J. Ginsberg, J.
Shade, D. Fulk. The digital Michelangelo Project: 3D
scanning of large statues. In Siggraph 2000, 2000,
131～144
[20] Stanford digital Formae Urbis Romae project,
http://formaurbis.stanford.edu/index.html
[21] P. Mueller, T. Vereenooghe, M. Vergauwen, L. Van Gool,
M. Waelkens. Photo-realistic and detailed 3D modeling: the
Antonine nymphaeum at Sagalassos (Turkey). Computer
Applications and Quantitative Methods in Archaeology
(CAA): Beyond the artifact - Digital interpretation of the past.
[http://www.vision.ee.ethz.ch/~pmueller/
documents/caa04_pmueller.pdf, accessed Mar. 2005]
[22] Qin Lian, Di-Chen Li, Yi-Ping Tang, Yong-Rui Zhang.
Computer modeling approach for a novel internal architecture
of artificial bone. CAD Computer Aided Design, 2006,
38(5): 507 ～514
[23] W. Sun, B. Starly, J. Nam, A. Darling. Bio-CAD modeling
and its applications in computer-aided tissue engineering.
Computer-Aided Design, 2005, 37: 1097 ～1114
[24] Wei Sun, Pallavi Lal. Recent development on computer aided
tissue engineering— a review. Computer Methods and
Programs in Biomedicine, 2002, 67: 85～103

Xi-Dao LUAN born in 1976. He is

currently a doctor candidate in
Multimedia R&D Center, National
University of Defense Technology. His
research interests include computer
graphics, virtual reality, computer vision
and multimedia.

Yu-Xiang XIE born in 1976. She

received her Ph.D in Computer Science
from National University of Defense
Technology in 2004. Now her research
interests include multimedia information
system and signal processing.

Long YING born in 1977. He is

currently a doctor candidate in
Multimedia R&D Center, National
University of Defense Technology. His
research interests include computer
graphics, virtual reality and computer
vision.