APPLICATION OF A THREE-DIMENSIONAL COLOR ...3D laser scanner uses three different Figure 1....

Palaeontologia Electronica http://palaeo-electronica.org

APPLICATION OF A THREE-DIMENSIONAL COLOR LASER SCANNER TO PALEONTOLOGY: AN INTERACTIVE MODEL OF A JUVENILE

TYLOSAURUS SP. BASISPHENOID-BASIOCCIPITAL

Patrick D. Lyons, Marc Rioux,and R. Timothy Patterson

Patrick D. Lyons, Department of Earth Sciences, Carleton University, Ottawa, Ontario K1S 5B6, CanadaMarc Rioux, Visual Information Technology Group, Institute for Information Technology, National Research Council, Ottawa, Ontario, K1A 0R6, CanadaR. Timothy Patterson, Department of Earth Sciences, Carleton University, Ottawa, Ontario K1S 5B6, Can-ada

ABSTRACT

Three-dimensional (3D) modeling has always been an important part of paleonto-logical research and interpretation though digital reproductions of fossils are a recentphenomena. A highly accurate, interactive, 100 µm resolution, 3D, digital model of afossilized basisphenoid-basioccipital from a juvenile Tylosaurus sp. mosasaur wasgenerated using a 3D laser scanner and manipulated using VRML and InnovMetricpolygon files. This 3D model supports varying levels of magnification depending on theinitial scan resolution and the amount of post-production polygon reduction. The gener-ation of these 3D models is relatively simple because the software and technology fortheir generation is relatively mature. At present, complex 3D models require powerfulcomputers in order to manipulate their computer graphic substructures. But, as com-puter technology improves, digital 3D scanning could prove invaluable for creating andsharing virtual copies of fossil material.

Key Words: Mosasaur, three-dimensional (3D), model, virtual reality, VRMLCopyright: Society for Vertebrate Paleontology, 15 November 2000Submission: 28 March 2000, Acceptance: 3 November 2000

INTRODUCTION

Increasingly, paleontologists are ableto exploit technology to aid in visualizingextinct life. For example, three-dimen-sional (3D) computed tomography (CT)techniques have been applied to deter-mine the external form of embedded fos-sils (Torres 1999), as well as to develop a

digitally rendered endocast model of aTyrannosaurus rex (Brochu 2000). Mod-els of conodonts have also been gener-ated by applying Virtual Reality MarkupLanguage (VRML) to predict the geometryof bedding plane arrangements (MacRae1995). Finally, a combination of QuickTimeVR and scanning electron microscopic

Lyons, Patrick D., Marc Rioux, and R. Timothy Patterson, 2000. Application of a Three-Dimensional Color Laser Scanner to Paleontology: an Interactive Model of a Juvenile Tylosaurus sp. Basisphenoid-Basioccipital. Palaeontologia Electronica, vol. 3, issue 2, art. 4: 16pp., 2.04MB. http://palaeo-electronica.org/2000_2/neural/issue2_00.htm

PATRICK D. LYONS, MARC RIOUX, & R. TIMOTHY PATTERSON: TYLOSAURUS 3D LASER SCAN

methods have been utilized to imagemicrofossils (Lyons et al. 1998). Three-dimensional modeling is not new to thefield of paleontology though.

Alcide d’Orbigny in 1826 (d’Orbigny1843) made some of the first 3D models offist-sized reproductions of microscopic for-aminifera. However, the most famousearly models were certainly the life-sizeinterpretations of Iguanodon, Hylaeosau-rus, Megalosaurus, Plesiosaurus andIchthyosaurus sculpted by BenjaminHawkins in 1853 for the International Exhi-bition. After the exhibition closed the dino-saur models were subsequently moved toSydenham Park in South London (Spald-ing 1993).

Traditionally, if a fossil important to thepaleontological community is discovered,the resultant research publication is usu-ally accompanied by appropriate illustra-tions. If this work is built upon, otherresearchers will either have the specimensent to their institution or will visit the insti-tution where the specimen is archived toexamine the specimen. Although tradi-tional modelling techniques (e.g., plasteror fiberglass casting) will undoubtedlyremain popular and relevant for some timeto come, there are real advantages inusing digital methodologies for researchpurposes. The benefit of utilizing a digitalmodel versus the actual fossil, or a tradi-tional reproduction, is the potential toshare and carry out research on fossilswith colleagues over great distancesquickly and cheaply with no danger to theoriginal material. Shape analysis, as wellas soft and hard tissue reconstruction canall be done easily within the digital realm.Hard copies of the digital model can alsobe produced using rapid prototypingmachines (mechanical devices used toturn 3D computer-generated designs intoproduction prototypes). These instrumentspermit the production of highly accuratemodels with a level of detail and accuracy

far exceeding those typical of traditionalcasts (Beraldin et al. 1997).

Other uses of high-resolution 3D laserscanners, as applied to fossil material,include the ability to provide external sur-faces (e.g., biomedical application such asimaging and reconstruction of brains. Wal-lace 1999) and creation of a collection ofdigital reconstructions that would allow fora comparison of surface structures acrosssimilar species.

Three-dimensional laser scannershave been in limited use since their devel-opment by the National Research Councilof Canada in 1981. However, it is onlywithin the last three years that these scan-ners have become more prevalent andaccessible to researchers and engineersthrough commercial development of thetechnology. As high-resolution (<100µm)laser scanners become commerciallyavailable, they represent a unique oppor-tunity to image fossil material. Three-dimensional laser scanners have alreadybeen applied in such diverse applicationsas documenting archaeological artifactsfrom The Canadian Museum of Civilization(Rioux 1994), documenting archaeologicaldigs (Boulanger et al. 1998), using reverseengineering (a process by which an objectis scanned in three dimensions and thenreconstructed physically with either rapidprototyping machines or automatedlathes. Godin et al.1996), and determiningthe authenticity of artwork (Baribeau et al.1992).

Work by other paleontologicalresearchers has used less sophisticated3D scanning technology, including low-resolution (>500µm) point scanners to pro-duce an animated Triceratops sp. in loco-motion studies (Andersen et al. 1999).Even lower-resolution laser scanners havebeen used in modeling the morphology ofa variety of vertebrate fossil material(Chapman 1997). To the best of our knowl-edge, though, high-resolution laser scan-

2


ners have not been applied topaleontological research.

In this article we present and explore anew technique for the 3D modeling ofmacrofossil material from a mosasaur.This technique utilizes a high-resolution3D laser scanner to capture a series ofpositional coordinates (x, y, z).

Mosasaurs are an extinct group ofmarine reptiles that were most diverse andabundant through the latter half of the Cre-taceous, and like other groups of marinereptiles became extinct at the K-T bound-ary (Russell 1967). They reachedimmense sizes, up to 10 meters in length,and are well-known from many recoveredspecimens, but relatively few juvenileshave been found (Caldwell 1996). Frombraincase material of a mosasaur in thecollections of the Canadian Museum ofNature, the basisphenoid-basioccipital(#51259) of a juvenile Tylosaurus sp. wascompletely imaged by the 3D laser scan-ning technique at a resolution of 100 µm.The basisphenoid-basioccipital is locatedat the base of the skull and is one of sev-eral bones that form the braincase in ver-tebrates. These bones support and protectthe brain.

Although paleontologists often dependheavily on the physical features wheninterpreting specimens, several questionsmust be answered if this technique is toprove viable in providing accurate digitalmodels fit for paleontological study. Thus,primary aims of this paper are to deter-mine whether a digital model scanned at100 µm is accurate enough for paleonto-logical study and whether available com-pression methods used to make suchmodels more easily accessible with desk-top computers reduce the scientific valueof the digital model by obscuring or evendeleting important features.

METHODS AND MATERIALS

A total of six bones from the braincaseof a juvenile Tylosaurus sp. were avail-able for three-dimensional laser scanning;the right and left prootic, the right quad-rate, the supraoccipital, the parietal, andthe basisphenoid-basioccipital. Thebasisphenoid-basioccipital (Fig. 1) waschosen for this initial evaluation becauseof its complex surfaces and its importancein revealing the positions and paths of cra-nial nerves.

The Institute for Information Technol-ogy (a division of the National ResearchCouncil of Canada NRCC) developed the3D laser scanner (Fig. 2) that was used.This scanner is able to generate extremelyaccurate scans at resolutions of as little 10µm. Scanning at such high resolutionsrequires significant time for both scanningthe object because the laser must physi-cally travel slower over the object andusing substantially more computer pro-cessing time in order to generate the 3Dmodel. Previous analysis of the accuracyof the 3D laser scanner used in this studyfor industrial prototyping purposes indi-cates that distortion-free models can begenerated down to a resolution of 10 µm(Beraldin et al. 1997).

Even a 50 µm-resolution scan allowsfor extremely detailed and accurate repro-ductions of objects. But even at this reso-lution it is done at the expense ofgenerating very complex polygonal mod-els with numerous individual polygons.Due to the complex surfaces present onthe basisphenoid-basioccipital, we esti-mated that scanning the bone at a resolu-tion of 50 µm would generate a modelcomposed of over ten million polygons. Asone might guess, models composed ofsuch large numbers of polygons causesignificant problems for present-day com-puter technology and generally requireprohibitively expensive technology to gen-erate. For this reason, prior to carrying out

3


a full scan of the basisphenoid-basioccipi-tal, a series of test scans were carried outat resolutions at both 50 µm and at 100µm on a small 1 cm2 area. Based on thesetests, it was determined that a resolutionof 100 µm produced satisfactory resultsand manageable polygon counts.

Having determined a satisfactory scanresolution, the basisphenoid-basioccipitalwas scanned 30 times in a variety of orien-tations to allow all surfaces to be exposedto the laser. The information captured bythe laser scanner was compiled on a Sili-con Graphics workstation. These 3Ddatasets were then imported into a soft-ware package developed by InnovMetricSoftware Inc. called PolyWorks/Modelerversion 5.0. Using the automatic alignmenttechnology built into PolyWorks/Modeler,the multiple datasets from the 30 scans indifferent coordinate systems were unifiedinto a single coordinate system, formingthe 3D surface. PolyWorks/Modeler'shigh-precision alignment algorithm allowsunrestricted movement of either the objector the digitizer to measure the entireshape of the object without any externalreference (Beraldin et al. 1997).

The 3D laser scanner employed atNRCC is also able to capture color infor-mation for each positional coordinate.Once this information is passed to Poly-works/Modeler, a texture map is generatedand applied to the digital model. A noteshould be made about color. The NRCC3D laser scanner uses three different

Figure 1. Three-dimensional, color, laser-scanned image of a basisphenoid-basioccipital of a juvenile Tylosaurussp. 100 µm scan resolution; Image from 800,000 polygon model.

Figure 2. Laser unit of 3D scanner developed by thedivision of Visual Information Technology, NationalResearch Council Canada.

4


wavelength (red, green, and blue) lasersto register accurate data on color reflec-tance (Soucy et al. 1996). As the lasers donot depend on ambient light for colordetermination, the texture maps generatedfrom the color information are accurate(Beraldin et al. 1997). However, there isvariation in the color information when themodel is displayed on cathode ray tube(CRT) or liquid crystal display (LCD) moni-tors or when images are printed. This vari-ation is due to limitations and individualvariations in how different computers’video cards and CRT-LCD monitor displaycolor information (Fraser 1998).

The completely assembled 100 µmresolution digital model was composed ofover 3 million polygons and totaled over

76 megabytes (MB) in size. To provide anindication of how file sizes balloon withincreasing resolution, the 1 cm2 test areascanned at a 50 µm resolution alone. Thisresulted in a model composed of 870,000polygons and was 23.5 MB in size. In con-trast, the same area scanned at 100 µmproduced an 11.75 MB file comprised of235,000 polygons. The completed modelswere saved as an InnovMetric polygon file(.pol). This proprietary file format is effi-cient and preserves texture information aswell as the coordinate system that formsthe basis for the model.

It was necessary to perform a seriesof polygon reductions to diminish the com-plexity and storage size of the model andto allow for interpretation and viewing of

Figure 3. Interactive polygon reduced (from 3 million to 50,000 polygons) VRML model of a juvenile Tylosaurus sp.basisphenoid-basioccipital. The overall length of the fossil is 113 mm.

Author's note: The preferred method for viewing the three-dimensional model is with Virtual Reality Modeling Language (VRML) software. However, as the VRML model is a large download (15 MB), is extremely demanding on desktop computers, and it requires a fast computer (minimum 400 MHz Pentium II (or equivalent) or a G3 processor, 64 MB RAM, and 8 MB of video RAM), a QuickTime VR version is provided. Viewing the VRML requires a VRML plug-in a format that is more fully supported on PCs than Apple Macintosh computers. The recommended VRML plug-in for Macintosh is WorldView from Intervista or the Macintosh version of the Cortona VRML plugin from Parallel Graphics. PC users should use the PC version of the Cortona plugin. Corona VRML: http://www.parellelgraphics.com/. WorldView from Intervista: http://tucows.apollo.lv/mac/plugmac.html.

5


the model on typical desktop computersystems. The completed 100 µm resolu-tion model described above was reducedto three models composed of 800,000 (21MB), 100,000 (2.6 MB), and 50,000 (1.3MB) polygons, respectively. As part of thisevaluation, the polygon count for the 1cm2 test area scanned at 50 µm resolutionwas also reduced from 870,000 polygonsto 50,000 polygons (1.3MB). Prior to thisreduction, still images of both the com-plete 3 million polygon model (100 µmscan resolution) and the 1 cm2 areascanned at 50 µm resolution were ren-dered and saved in JPEG file format.

All reduced polygon files were con-verted from the proprietary InnovMetric fileformat to VRML (ver. 2.0) (Fig. 3). TheVRML format is a standard text-based 3Dfile format that allows for viewing of thefiles over the World Wide Web. Unfortu-nately as this file format is text based, filesizes increase dramatically.

Both the original .pol files and theVRML files were transferred to CarletonUniversity from NRCC for manipulationand analysis on both Apple Macintosh andIBM compatible PCs. The VRML files wereviewed and evaluated using a software

package from Auto-des-sys Inc. entitledForm-Z (ver 3.1.4). This software applica-tion allowed for relatively rapid display ofthe 3D models and corrected the problemsencountered when using Intervista. UsingForm-Z, the texture map could beremoved and the model could be analyzedwithout color bias (Fig. 4). To decrease thefile size and to allow for greater access to

Figure 4. Three-dimensional, color, laser-scanned image of a basisphenoid-basioccipital of a juvenile Tylosaurus sp.One-hundred micrometer scan resolution with color information removed.

Figure 5. Interactive QuickTime VR object movie of ajuvenile Tylosaurus sp. basisphenoid-basioccipital.There are rendering artifacts present in the model visi-ble as fine lines delineating the wireframe, which sup-ports the model. These artifacts are not present in theVRML model and are a result of the software packageused to generate the QuickTime VR object movie. Theoverall length of the fossil is 113 mm.

Author's note:Viewing the QuickTime VR version of the model requires QuickTime 4.0. QuickTime is available for both Macintosh and PC platforms.

6


the model, a QuickTime VR object wascreated (Fig. 5) using Form-Z’s export to aQuickTime VR feature. Scanning thebasisphenoid-basioccipital at 100 µmrequired four hours, with an additional 3hours of postproduction time used to com-pile, assemble, and reduce the model.

RESULTS

Digital model accuracy

The 800,000 polygon digital modelscanned at 100 µm was compared to thephysical specimen by scaling on the com-puter screen and by using microscopictechniques respectively. The dimensions,geometry, and color were identical on thedigital model to that of the specimen. Fig-ure 6 is a composite image comparing the

Figure 6. Dorsal view of Tylosaurus sp. basisphenoid-basioccipital comparing traditional photography and 3D laserscanning. Color information is removed so that comparison can be done without color bias. Three-dimensional, 100µm resolution scan image from 800,000 polygon model.

7


digital model with a photograph of thephysical specimen. To the eye, even atincreased magnification, it was not possi-ble to differentiate the specimen and thedigital copy. As accuracy of the methodol-ogy has been previously rigorously testedunder demanding industrial conditions itwas not deemed necessary to quantify theverification process. However, visualexamination of two regions highlightsexamples of the accuracy of the digitalmodel. Figure 7 illustrates the vidiancanal; its structure can be seen easily inboth the model and the image of the phys-ical specimen. Figure 8 shows details ofthe abducens cranial nerve (VI) and thebranch of the internal carotid artery (ICB).The only limitation of digital models suchas shown is that they can only support alimiting amount of scaling (Fig. 9). The lim-itation of this scaling is related to two vari-ables; the scanning resolution andpolygon reduction (Beraldin et al. 1997).

Comparison of 50 µm and 100µm scan

resolutions

Figure 10 was scanned at 50 µm reso-lution, and Figure 11 was scanned at 100µm resolution. Both models support sub-stantial scaling without facets appearing inthe digital model. However, when the 100µm digital model is scaled to more than 5x,facets become visible on high angle sur-faces (Fig. 12). In contrast, the modelscanned at a 50 µm resolution supportsmuch greater scaling, to the point ofexceeding the resolution of the texturemap (Fig. 13).

Polygon reduction

As discussed earlier, to permit mostresearchers access to a version of thiscomplex digital model that can be manipu-lated, it was necessary to reduce the num-ber of polygons. Figure 3 is a digital modelcomprised of 50,000 polygons scanned ata 100 µm resolution. This represents a98% reduction in the number of polygonsfrom the initial 3 million polygon model and

Figure 7. High angle, 3D, 100 µm resolution scan image from 800, 000 polygon model illustrating structure of vidiancanal.

8


9

Figure 8. Dorsal view of both traditional photography and 3D laser scanning of Tylosaurus sp. basisphenoid-basioc-cipital highlighting the Cranial Abducens Nerve (VI) and the branch of the internal carotid artery (icb). Three-dimen-sional, 100 µm resolution scan image from 800,000 polygon model.


can only support a limited amount of scal-ing (1.5x) before details apparent in thephysical specimen are either obscured orabsent (Fig. 14).

DISCUSSION

One fundamental question that thisinvestigation sought to address waswhether a high-resolution, 3D, laser-scanned, digital model would be useful topaleontologists. Certainly a digital model

Figure 9. Magnification of 3D 100 µm resolution scan from 800, 000 polygon Tylosaurus sp. basisphenoid-basioccip-ital model illustrating faceting.

Figure 10. 1 cm2 area of basal tuber of Tylosaurus sp. basisphenoid-basioccipital scanned at 50 µm resolution.

10


allows for a new method of illustration. Butin some cases one must consider whethera digital model represents a viablereplacement for a physical specimen orcast?

Several areas of the 3-million polygondigital model were examined and com-pared to the specimen. For example, inthe case of the basisphenoid-basioccipitalpresented here, areas of interest to aresearcher examining a mosasaur brain-case would be blood supply to the brainand the cranial nervous system (Russell1967). Obviously, soft tissues are rarelypreserved (Fastovsky and Weishampel1996), but cranial nerves often left path-ways and foramens throughout the braincase as they traced their way from theextremities of the animal (e.g., sensoryorgans) to the various reception sites inthe brain (Hildebrand 1988). These fora-mens can be essential to researchersbecause they indicate the likely positionand structure of the brain (Hildebrand1988) and can provide important charac-ters for a phylogenetic analysis (Bell1997). As a result of alteration during the

fossilization process, the illustratedbasisphenoid-basioccipital (Figures 3 and5) shows some deformation and compres-sion along the lateral margins, slightlyobscuring one of the foramen of theabducens nerve (cranial nerve VI; Fig. 8).However, the structure is clearly visible,even scanned at 100 µm resolution.

Blood circulation to the brain also pro-vides important clues to phylogenetic rela-tionships, with the vidian canal likelycontaining the internal carotid artery (Rus-sell 1967; seen clearly in Fig. 7). Thebranch of the internal carotid artery leavesthe vidian canal and passes mediallythrough the basisphenoid (Fig. 8). In somemosasaur clades the internal CarotidArtery became significantly enlarged so itis of phylogenetic importance to recon-struct this vessel.

Comparison of 50 µm and 100 µm scan resolution

Would a 100 µm scanning resolutionbe sufficient to produce an accurate digitalmodel for paleontological study, or is a 50µm scanning resolution a requirement?

Figure 11. 1 cm2 area of basal tuber of Tylosaurus sp. basisphenoid-basioccipital scanned at 100 µm resolution.

11


12

Figure 12. 1 cm2 area of basal tuber of Tylosaurus sp. basisphenoid-basioccipital scanned at 100 µm resolution indi-cating faceting.

Figure 13. 1 cm2 area of basal tuber of Tylosaurus sp. basisphenoid-basioccipital scanned at 50 µm resolution illus-trating pixilation of texture map. Note that the facets of the underlying geometry are not visible.


The answer to this rhetorical question isimportant because scanning at higher res-olution generates a higher polygon countmodel and larger file sizes and requires asignificantly more powerful computer to

manipulate. As discussed above, whethera 50 µm or 100 µm resolution scan is suffi-cient for research purposes will depend onthe individual researcher’s interests. As anexample, a small area of the basal tuber

Figure 14. Lateral view of both traditional photography and VRML model (generated by a polygon reduction from 3million to 50,000) illustrating details and ridges apparent in physical specimen are either obscured or absent.

13


(Fig. 10) scanned at 50 µm was comparedto the same region at a 100 µm scan. Thebasal tuber of the basisphenoid is ridgedand pitted, likely from a rich supply ofblood through a network of blood vesselssupplying a cartilage covering (Russell1967). When the pits and ridges of thebasal tuber are magnified (Fig. 13), theinternal structure of the ridges are clearlyseen at 50 µm scanning resolution. Unfor-tunately at 100 µm scanning resolution thepits and ridges are faceted at the equiva-lent magnification. Clearly, if a researcheris interested in the internal structures ofsmall features such as the basal tuber, orother structures smaller than 100 µm, ascan resolution of 50 µm is required. How-ever, it is likely that, for a majority ofresearchers and for most applications, a100 µm scan resolution would be satisfac-tory because it supports artifact-free mag-nifications of up to 5x.

As laser scanning technologymatures, as the process becomes increas-ingly automated, and as personal comput-ers become faster, cheaper, and morecommon, scans at 50 µm resolution willbecome more practical permitting moreinformation to be presented in a 3D model.

Polygon Reduction

While computer speed has increaseddramatically in recent years (Moore 1997),most desktop computers have yet toachieve performance levels suitable formanipulating 3D digital models comprisedof hundred of thousands of polygons inreal time. Modern Silicon Graphics work-stations allow real-time manipulation ofdigital models composed of millions ofpolygons. Unfortunately, the prohibitivecost of these high-end graphics computersmakes them unavailable to most research-ers. To permit most readers access to aversion of this complex digital model thatcan be manipulated, it was necessary toreduce the number of polygons to a more

manageable number. ‘Manageable num-bers’ of polygons for desktop computersare dependent on the speed of the proces-sor, the speed of the video card, theamount of video memory, the amount ofrandom access memory available, and thefile type. At this time 50,000 to 800,000polygons seems to be the limit for themost recent desktop machines (e.g., IntelPentium III, AMD Athalon, and MotorolaG3/G4 processors).

Consideration of the differencesbetween 3D file types is critical whenassessing the ability of desktop computersto display a complex model composed ofmany polygons. Some file types supporttexture maps and multiple light sourcesand allow for compact file sizes, but arenot widely viewable. In contrast, other filetypes do not support texture maps and areinefficient, but are well-supported and eas-ily read by a variety of operating systems(Macintosh, PC, and Unix). An example isInnovMetric’s Polygon file type versusVRML. With the Intel Pentium III computer,used for this research, with 384 MB ofRAM, an InnovMetric polygon file com-posed of 800,000 polygons is easilymanipulated in near real-time, while thesame model must be reduced to 50,000polygons to achieve the same responsewhen viewed using VRML. The benefit ofusing VRML though is that it is a standardfile type and can be read by many differenttypes of software. These VRML files willalso likely still be readable for the foresee-able future because the files are textbased. For these reasons, VRML waschosen as the preferred file type for thisstudy even though it is not nearly as effi-cient as an InnvoMetric polygon file for-mat. Nevertheless, it should be noted thatVRMLs inefficiency required that the num-ber of polygons be reduced from 3 millionpolygons to 50,000 polygons; a reductionof 98%. While this allows for the model tobe displayed on fast desktop computers,

14


an assessment must be made as towhether any paleontological valueremains in such a substantially reducedmodel. An algorithm present in PolyWorks/Modeler that removes redundant informa-tion carried out the process by which thenumber of polygons was reduced. Resultssuggest that if the 50,000 polygon digitalmodel is displayed at its original (1x) size,no apparent difference can be notedbetween the physical specimen and thedigital model (Fig. 3). However, as the dig-ital model is scaled upwards (simulatingexamination of the physical specimen atincreased magnification), differencesquickly become apparent. At a scaleequivalent to 1.5x and upwards, the finedetails and ridges apparent in magnifiedviews of the physical specimen are eitherobscured or absent (Fig. 14). Thus, poly-gon-reduced models only have limitedvalue to researchers, for example in rolessuch as informal consultation with col-leagues or popular science.

Polygon-reduced models potentiallydo have a role in paleontological researchin other applications such as virtual recon-struction though. For example, if the fiveremaining braincase bones (right quad-rate, right and left prootic, parietal andsupraoccipital) of the juvenile Tylosaurussp. were also scanned at 100 µm resolu-tion and then assembled to create a virtualreconstruction of the brain case, theresultant model would likely comprise over12 million polygons and would be inexcess of 200 MB. In order to manipulatesuch a reconstruction, the number of poly-gons would have to be reduced by anorder of magnitude or more depending onthe researcher’s hardware and software.The resultant reduced model would still beuseful for various applications includingestimation of brain morphology and vol-ume. One can imagine the number of poly-gons present in a complete whole skeletonreconstruction though; it could conceivably

number in the billions leaving this applica-tion in the realm of science fiction for thenext few years.

CONCLUSIONS

Three-dimensional laser scanners areable to generate highly accurate and pow-erful digital reconstructions that, in turn,are able to support significant levels ofmagnification. Digital models can, in somecases, replace the physical specimendepending on the level of detail sought bythe researcher. Scanning at 100 µm reso-lution is suitable for most vertebrate pale-ontological research if the researchinvolves analysis of structures larger than100 µm. If the structures of interest aresmaller than 100 µm, a 50 µm scanningresolution must be used. Although makingmodels easier to manipulate, polygoncompression reduces the level of detailpresent in the model and should not beused for detailed paleontological study.Development of digital models will make iteasier for collaboration over the WorldWide Web, making it easier to protectvaluable specimens. Other uses couldinclude the modeling of internal structures.This technique would also allow for high-precision realignment of serial sectionsand the development of a digital modellibrary for comparison of similar speci-mens. With the development of more pow-erful desktop computers in the next fewyears, it will be possible for the develop-ment of virtual reconstructions of importantsystems or complete skeletons that areeasily manipulated. As the technologycontinues to mature, the potential existsfor specimens to be archived digitally asthree-dimensional models that could thenbe stored in a database for rapid retrievaland referencing.

ACKNOWLEDGEMENTS

We thank the Visual Information Tech-nology division of the National Research

15


Council of Canada for use of the three-dimensional laser scanner and theirincredible expertise and patience, M.Caldwell of the Canadian Museum ofNature for arranging the loan of the fossilmaterial and initial conversations, M. Getzof the department of Industrial Design atCarleton University for help with Form-Zand file conversions, A. Webb for the loanof the IBM- compatible computer and C.Schröder-Adams and N. Saumure foradvise and support. This project wasfunded by NSERC Research GrantOGP0041665 to R. T. Patterson.

REFERENCESAndersen, A. F., Chapman, R. E., Kenny, K., and Lars-

son, H.C.E., 1999. Animation of 3-D Digital Data: TheWalking Triceratops. Abstracts 59th Annual Meet-ing of the Society of Vertebrate Paleontology,Denver, 19(3):29A.

Baribeau, R., Cournoyer, L., Godin, G., and Rioux, M.,1996 Colour Three-Dimensional Modelling ofMuseum Objects, Imaging the Past, Electronic Imag-ing and Computer Graphics in Museum and Archae-ology. British Museum, Occasional Paper 114:199- 209.

Baribeau, R., Rioux, M. and Godin, G., 1992. RecentAdvances in the use of a Laser Scanner in the Exam-ination of Paintings, Proceedings of Restoration'92, Amsterdam, 69-73.

Bell, G. L., 1997. Phylogenetic revision of the Mosasau-roidea, p.281-332. Callaway, J. M. and Nicholls, E. L.(ed.), Marine Reptiles Academic Press, Toronto.

Beraldin, J.-A. Cournoyer, L. Rioux, M. Blais, F. El-Hakim, S. F. and Godin, G., 1997 Object model cre-ation from multiple range images: acquisition, cali-bration, model building and verification.Proceedings of the International Conference onRecent Advances in 3-D Digital Imaging and Mod-eling Ottawa, NRC 40169:326-333.

Boulanger, P., Taylor, J., El-Hakim, S. and Rioux, M.,1998 How to Virtualize Reality: An Application to theRe-creation of World Heritage Sites, VSMM98, Inter-national Society on Virtual Systems and Multime-dia, Gifu, Japan 1:39 – 45.

Brochu, C., A., 2000. A digitally-rendered endocast forTyrannosaurus rex. Journal of Vertebrate Paleon-tology, 20:1-6.

Caldwell, M.W. 1996. Ontogeny and phylogeny of themesopodial skeleton in mosasauroid reptiles. Zoo-logical Journal of the Linnean Society 116:407-436.

Chapman, R. E., Rasskin-Gutman, and Hunt, G., 1997.Using computer modeling of morphology for explor-ing functional and evolutionary models Abstracts57th Annual Meeting of the Society of VertebratePaleontology, Chicago, 17(3):37.

d’Orbigny, A., 1843. Modeles de Foraminiferes vivanset fossils. Imprimerie de Cosson, Paris.

Fastovsky, D. E., and Weishampel, D. B., 1996. TheEvolution and Extinction of the Dinosaurs. Cam-bridge University Press, New York.

Fraser, B, 1998. The Power of ColorSync Color Match-ing. Macworld June 1998 105-109.

Godin, G., Soucy, M., and Boulanger, P., 1996. Rangeimage integration for direct replication of objects.SPIE Proceedings On Rapid Product Develop-ment Technologies, Boston, 2910:34-44.

Hildebrand, M., 1988. Analysis of Vertebrate Struc-ture. John Wiley & Sons, Inc., Toronto.

Lyons, P. D. and Head, L, 1998. QuickTime VR: A power-ful new illustrative tool for micropaleontologicalresearch. Palaeontologia Electronica, 1(2): 12p., 219KB. http://www-odp.tamu.edu/paleo/1998_2/lyons/issue2.htm

MacRae, A. R. and Henderson, C. M., 1995. 3-dimen-sional models of the Scottognathus conodontassemblage. Abstracts Proceedings of the 1995GAC/MAC convention.

Moore, G. E., 1997. The Microprocessor: Engine of theTechnology Revolution. Communications of theACM 40(2):113-114.

Rioux, M., 1994. Digital 3-D Imaging: Theory and Appli-cations. SPIE Proceedings, Videometrics III,Inter-national Symposium on Photonic and Sensorsand Controls for Commercial Applications, Bos-ton, 2650, 2-15.

Russell, D. A., 1967. Systematics and Morphology ofAmerican Mosasaurs. Peabody Museum of NaturalHistory Bulletin 23:1-240.

Soucy, M., Godin, G., Baribeau, R., Blais, F. and Rioux,M. 1996. Sensors and Algorithms for the Construc-tion of Digital 3-D Colour Models of Real Objects,ICIP-96 Proceedings, 1996 IEEE InternationalConference on Image Processing, Lausanne, Swit-zerland, 2:409-412.

Spalding, D. A. E., 1993. The Dinosaur Hunters. KeyPorter Books Ltd. Toronto.

Torres, A. M., 1999. A three-dimensional CT (CAT) scanthrough rock with Permian alga Ivanovia tebagaen-sis. Journal Of Paleontology 73:154-158.

Wallace, 1999. Medical Diagnostics, Laser FocusWorld, v. 35 (1):1

16

Date post:	18-Sep-2020
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

APPLICATION OF A THREE-DIMENSIONAL COLOR ...3D laser scanner uses three different Figure 1....

Documents