Thule peoples are the cultural and biologicalancestors of contemporary Inuit and Eskimo

groups of the North American Arctic and Green-land. By the late 12th or early 13th century,Thule groups had expanded eastward from theBering Strait region into the Canadian Arctic.Unlike northwestern Alaska, the coastlines of theEastern Arctic were largely devoid of driftwood.Consequently, the Thule people used whaleboneto construct the roofing frameworks of theircoastline winter houses. They erected the roofframework over a house pit furnished with a flag-stone floor, raised sleeping platform, kitchen, andstorage areas; and covered the roof frame withhide and a thick layer of turf, moss, and snow.1,2

Archaeologists know little about how these enig-matic houses were constructed because few existintact. Reconstructing a 3D model of a Thulehouse can provide new insights into how theThule people constructed and used thesedwellings.

Our reconstruction of a Thule whalebonehouse provides a good case study of laser scan-ning’s use on an object of complex geometry.The reconstruction process would have been dif-

ficult, if not impossible, to resolve using 2D draw-ings because manual drafting or 2D computer-aided design (CAD) can’t easily solve a 3Dstructural system based on complex skeletal ele-ments, such as a whale’s mandibles, cranium,and maxillas.

Early history of the projectArchitectural research of Thule whalebone

houses began in the summer of 1994, when PeterDawson and a team of archaeologists from theUniversity of Calgary constructed accurate mapsof 23 whalebone houses at the Deblicquy site(QiLe-1, shown in Figure 1a), an exceptionallywell-preserved Thule archaeological site onBathurst Island, Nunavut.3,4 We placed a grid of3 × 3-meter units over each house mound. Wethen carefully measured and drew each gridunit’s structural and architectural components toscale, as Figure 1b shows. When we returnedfrom the field, we drew plans of the site usingAutoCAD software.

The value of laser scanning In 2003, we began to explore strategies for cre-

ating 3D computer reconstructions of the Thulewhalebone houses based on the earlier field studyresults. The ultimate goal was to test the designof houses made of bowhead whale elements—found at archaeological sites. We could constructan accurate model only if we had the major struc-tural elements’ geometry (mandibles, cranium,maxillas, and ribs). Fortunately, the New EnglandAquarium in Boston has a mounted specimen ofa North Atlantic right whale skeleton, as Figure2a shows. Although the North Atlantic rightwhale (Eubalaena glacialis) is smaller than thebowhead whale (Balaena mysticetus), its skeletalmorphology is similar. Thus, by scaling appro-priately, we could use the North Atlantic rightwhale skeleton to model bones found during thearchaeological excavation.

78 1070-986X/06/$20.00 © 2006 IEEE

Multimedia at Work Qibin SunInstitute for Infocomm Research

Richard Levy andPeter Dawson

University ofCalgary

Reconstructing a Thule WhaleboneHouse Using 3D Imaging

In this article, the authors share with us an interesting use for the latestlaser-scanning-based 3D imaging technology for reconstructing a ThuleWhalebone House. All major components (such as date capturing, datamodeling, display, and interaction) are covered to demonstrate howarchaeological research can benefit from this new technology for the pur-poses of testing and education.

Although this article focuses on building the skeletal models of a whale,it also provides a stepping-stone to the researchers, engineers, architects,and archaeologists who are interested in virtual reality. Interactive digitalmedia is becoming one of the hot areas in the multimedia community,which aims to provide users with an immersive experience while consum-ing media.

—Qibin Sun

Editor’s Note

In the last decade, laser-scanning technologyhas emerged as an important tool in historicpreservation and archaeological investigation.Although lasers aren’t new technology, their useas measuring devices is relatively recent.5-8 Priorto laser scanning’s development, creating a 3Dmodel of a site or object required numerous fieldmeasurements and many weeks of data collectionusing optical theodelites, tapes, and calipers.5,7-11

In creating a model of the North Atlantic rightwhale skeleton, a Cyrax 2500 laser scanner offeredgood accuracy (5 millimeters) at ranges of 50meters.12 Reconstruction involves converting mil-lions of points into an optimized mesh, whichtakes several steps. Figure 2b shows the meshmodel we created from the scanning data, and Fig-ure 3 shows a flowchart of the process used to cre-ate the various parts of the model needed for thereconstruction and testing.

The exact steps encountered in the translationto mesh can vary depending on the applicationused for modeling, although the general issueswill remain the same. In this case, we imported apoint file from Cyra’s Cyclone (PTX format) intoInnovmetric’s Polyworks (http://www.innov-metric.com) to create the mesh files needed formodeling. The last step of mesh optimization canbe critical in scenes where high-count polygonmeshes must be reduced to low-count models inorder that today’s PCs can render in real time (seeFigures 2b and 3).

Modeling in virtual spaceThe electronic reconstruction process wasn’t

unlike building the actual physical structure. Thefirst step involved importing the 2D CAD file ofinformation collected at the Deblicquy site in1994.3,4 The plan for the largest and best-pre-served house (Figure 1a) served as the basis for 3Dreconstruction. This CAD data provided essentialinformation for the reconstruction, including thesubterranean pit’s topography, extent, andshape, which represent the dimensions of theenclosed space.

The list of bone types and sizes was also essen-tial to this reconstruction. This informationhelped us scale the individual elements builtfrom the laser-scanning data. Bones used in theoriginal structure included the mandible, maxil-la, cranium, ribs, scapulas, and selected vertebrae.We extracted these elements from the skeletalmodel using Polyworks. We used Autodesk’s3DStudioVIZ for the actual modeling process,applying texture maps based on digital images to

mesh forms to give a realistic appearance. The second step involved extracting the pit

from the topography using average depths andpit outlines in the original CAD file. We built aflagstone floor and elevated sleeping platformusing virtual rocks whose shapes, sizes, and colorwe determined using actual rocks measured atthe site. To begin the reconstruction process, weplaced the major construction elements—crani-um, mandibles, and maxillas—in their approxi-mate locations found on the site. One issue that

79

(a)

(b)

Figure 1. Thule

whalebone house

(QiLe-1) on Bathurst

Island, Nunavut:

(a) photograph of the

archaeological site and

(b) computer-aided

design (CAD) drawing.

80

Multimedia at Work

emerged early in the reconstruction process wasdetermining the bones’ exact locations. Overtime, bones can move from their original posi-tions, such as when a structure collapses. Giventhat the mandible or cranium’s spring point iscritical to the reconstructed form, the bone’s ini-tial location is only a best guess, which we refinethrough subsequent testing.

Our first hypothesis was to place the majorsupports along the pit’s perimeter. However, thisarrangement made it difficult to enclose the inte-rior space. We estimated that all of the boneswould have to be 10 percent longer than theiractual measurements to effectively create a tent-like structure.

We then placed the bones against the pit’s

Figure 2. (a) North Atlantic right whale skeleton at the New England Aquarium. (b) Mesh model of the New England Aquarium’s

specimen created from laser-scanning data.

(b)(a)

Laser-scanning datacyclone

Input format: IMPOutput format: PTX

Media/Virtual reality

Virtual reality

Virtools

SolidWorks/Cosmo

Input format: IGES

Multiframe 4D

Polyworks

Input format: PTXOutput format: OBJ

Mechanical desktop

Input format: IGESOutput format: IGES

Polytrans

Input format: OBJOutput format: 3DS

Media products:

AVI, MOV,QTVR,TIFF, JPG

3D Studio

Input format: OBJOutput format: MAX

Figure 3. Flowchart of

point-cloud data

processing.

walls, assuming the Thule builders would haveexcavated this subterranean area and lined it withflat stones. This arrangement results in more ver-tical elements than the first hypothesis. More ver-tical elements result in less applied force on theabutments and less splaying of the arch-like formsin these tent-like structures, as Figure 4 illustrates.Placing the major supports in a more verticalposition also reduces the elements’ flexure andcreates a house form with more interior space.

Although placing the bones on the pit’s insideperimeter solved the problem of spanning thesmaller of the two rooms (the cooking area), twoproblems emerged in crossing the larger space(the sleeping area).

❚ Bridging the larger space with the elementsfound on the site proved difficult given thedistance between the spring points of the cra-niums, mandibles, and maxillas.

❚ The skin, which acts as a membrane over thebone structure, can’t support heavy loads ofsod and snow alone. Without a ridgepolebetween the apexes of the domed spaces, thesod and snow’s weight would produce a sag inthe structure, potentially destroying itsintegrity.

It wasn’t uncommon for the Thule people tomove major bones from one site for use in con-struction at an adjacent site. Assuming bonesmight have been moved over time, we locatedseveral bones on adjacent sites of a size neededto enclose and span the spaces of DeblicquyHouse 8. The final design uses a mandible as aridgepole supporting the skin, sod, and snow’sweight in the saddle area connecting the twospaces (Figure 4).

To construct a whalebone house, Thulebuilders would have lashed ribs to the major sup-ports. Like gusset plates in a steel frame, theseribs helped create a more rigid structure. In aThule whalebone house, the ribs are particularlyimportant. Because the mandibles resting on thestructure’s floor act somewhat like pin connec-tions, without the ribs these elements wouldtend to rotate, causing instability in the entirestructure. Builders would also have used the ribsto create an entrance-way tunnel with doorsmade from the whale’s scapulas. A mesh com-pleted the model’s structure, representing thedraped skins and layered sod of an actual whale-bone house.

Prior to rendering images of the completedhouse form, we added lights of the intensity ofwhale-oil lamps to the interior space, as Figure 5illustrates. The renderings completed in Autodesk’s3DStudioVIZ exploit photometric-rendering algo-

81

Figure 4. Computer

reconstruction of the

Thule whalebone house

showing the ridgepole

design.

Figure 5. Computer reconstruction of the interior of a Thule whalebone house.

(a)

(b)

Multimedia at WorkIE

EE M

ulti

Med

ia

rithms that simulate light intensity and surfacelevels of reflectivity. Unfortunately, an estimateof a whale oil lamp’s illumination power onlygives us an approximate lighting solution. Futureresearch with whalebone lamp replicas will pro-vide empirical data on their luminescent charac-teristics and should generate more accuraterenderings of the interior space.

Structural stabilityTo go beyond the visual-inspection level, we

imported mesh files representing the subterraneanpit and bones into Dassault Systems’ Virtools (seehttp://www.virtools.com), an application for cre-ating virtual worlds. This approach lets us assignphysical properties to each structural element.Using the Havoc physics engine (a Virtools mod-ule), we assigned a mass to each whalebone and afriction coefficient to the bone and stone sur-faces. To simulate the lashing of the major ele-ments at each room’s apex, we created a series ofpivots or hinges to tie the structure together. Invirtual space, the structure seeks equilibrium afterbeing assigned physical properties and will eitherremain standing or will collapse under its ownweight. This approach let us refine the recon-struction by obtaining information on the struc-tural elements’ locations after they’ve reached anequilibrium state, as Figure 6 illustrates. Althoughonly a first step in a complex form’s structuralanalysis, adding a degree of physicality to the vir-tual objects can help archaeologists test a recon-struction’s validity in an interactive world.

Display and interaction Using virtual-world authoring, a research team

can create interactive worlds that they can distrib-ute on the World Wide Web or view in Cave Auto-matic Virtual Environments (CAVEs).9,13,14

Currently, we’re testing the use of a version of theThule whalebone house for teaching archaeology.Running the Virtools’ VRPack module, we can viewthe model with shutter glasses on a CRT display orproject it in a CAVE environment. This activestereo format depends on sequential frame buffer-ing of left and right views. At a minimum framerate of 60 Hz, the image will appear in stereo.

At the University of Calgary CAVE, we’re cre-ating the 3D effect using four projectors sup-ported by a cluster of four PC computers. EachPC in the cluster drives a single projection screen:left, right, front, and floor. A Dell Precision 650with Nvidia video cards (FX3000G) lets us displaythe complex geometry in real time. Interactivesound and atmospheric lighting all contribute tothe experience’s totality.

The opportunity to view the whalebone housein an immersive environment helps studentsunderstand the dwellings’ complex geometry andconsider various research hypotheses in a highlygraphical context using objects with virtual phys-ical properties. This helps them understand basicstructural principles and lets them explore con-nections between space, light, and culture.

One issue in the CAVE environment is thatinteraction is generally limited to a single user, mak-ing it more of a theater for those observing withouttrackers and other input devices. Although few col-lege campuses offer CAVEs, the ability to constructone from standard workstations will greatly expandthis technology’s future in research and education.In a museum environment, the real challenge increating experiences in these environments isinteraction with the virtual world.15 With decreas-ing hardware costs, the only barrier will be theprice of developing engaging virtual worlds, whichboth entertain and educate the public. MM

References1. M. Maxwell, Prehistory of the Eastern Arctic, Acade-

mic Press, 1985.

2. R. McGhee, Canadian Arctic Prehistory, Van Nostrand

Reinhold, 1978.

3. P. Dawson, “Interpreting Variability in Thule Inuit

Architecture: A Case Study from the Canadian High

Arctic,” American Antiquity, vol. 66, no. 3, 2001,

pp. 453-470.

4. M. Levy, P. Dawson, and C. Arnold, “Reconstruct-

Figure 6. Structural

analysis, before and

after achieving

equilibrium (ghosted

image).

82

ing Traditional Inuit House Forms Using Three-

Dimensional Interactive Computer Modeling,”

Visual Studies, 2004, vol. 1, pp. 26-35.

5. A.C. Addison, “Virtual Heritage—Technology in the

Service of Culture,” Proc. 2001 Conf. Virtual Reality,

Archaeology, and Cultural Heritage, ACM Press, 2001,

pp. 28-30.

6. M. Levoy et al., “The Digital Michelangelo Project:

3D Laser Scanning of Large Statues,” Proc. Siggraph,

2000, pp. 131-144.

7. K. Molenbrey, “Preserving the Past,” Computer Graph-

ics World, Sept. 2001, pp. 24-30.

8. P. Debevec et al., Estimating Surface Reflectance Prop-

erties of a Complex Scene under Captured Natural Illu-

mination, Univ. of Southern Calif. Inst. for Creative

Technologies, tech. report ICT-TR-06, 2004, pp. 1-11.

9. A.G. Gaitatzes, D. Christopoulos, and M. Roussou,

“Reviving the Past: Cultural Heritage Meets Virtual

Reality,” Virtual Reality, Archaeology,and Cultural

Heritage (VAST), 2001, pp. 103-109.

10. K. Nuyts et al., “From a Conservationist’s Point of

View,” Proc. Conf. Optical 3-D Measurement

Techniques V, Katholieke Universiteit Leuven, 2001,

pp. 179-186.

11. S.F. El-Hakim et al., “Detailed Reconstruction of

Large-Scale Heritage Sites with Integrated

Techniques,” IEEE Computer Graphics and Applica-

tions, vol. 24, no. 1, 2004, pp. 21-29.

12. M. Johansson, “Explorations into the Behavior of

Three Different High-Resolution Ground-Based Laser

Scanners in the Built Environment,” Proc. CIPA Int’l

Workshop Scanning for Cultural Heritage Recording,

2002; http://www.isprs.org/commission5/workshop.

13. A.C. Addison, “Emerging Trends in Virtual Heritage,”

IEEE MultiMedia, Apr.–June 2000, pp. 22-25.

14. E. Berndt and J. Carlos, “Cultural Heritage in the Mature

Era of Computer Graphics,” IEEE Computer Graphics

and Applications, vol. 20, no. 1, 2000, pp. 36-37.

15. E. Champion, “Applying Game Design Theory to

Virtual Heritage Environments,” Proc. Graphite Int’l

Conf. Computer Graphics and Interactive Techniques,

ACM Press, 2003, pp. 273-274.

Readers may contact Richard Levy at rmlevy@ucalgary.ca

and Peter Dawson at pcdawson@ucalgary.ca.

Contact Multimedia at Work department editor Qibin

Sun at qibin@i2r.a-star.edu.sg.

83