One Head, many Approaches – Comparing
3D Models of a Fossil Skull
Ilja KOGAN, Mirosław RUCKI, Maik JÄHNE, Daniel EGER PASSOS,
Tom CVJETKOVIC and Sascha SCHMIDT
In the frame of developing digitization standards for fossils, we are exploring the strengths
and weaknesses of different 3D imaging and replication approaches on the example of a Tri-
assic reptiliomorph amphibian skull. The holotype of Madygenerpeton pustulatum, a unique
and well-preserved tetrapod skull roof with a complex morphology, prepared from both the
dorsal and the ventral side, has been digitized using a range of structured-light scanners, a
laser scanner, computed micro-CT, and photogrammetry. Additionally, a series of 3D prints
has been produced based on one of the digital models. Both digital and analogue 3D models
are compared qualitatively and (semi)quantitatively, with the preliminary conclusion that a
considerably high accuracy of digitization and replication can be obtained with accessible
and user-friendly devices.
Ever since, petrified remains of vertebrates have attracted human attention, inspired myths
and legends, fascinated children and adults, were used as evidence for the Flood or for the
drift of continents. Much palaeontological information, elucidating the biology and interre-
lationships of extinct animals, can be derived from better-preserved vertebrate fossils, which,
however, are unlikely enough to ‘survive‘ millions of years in order to be discovered and,
thus, are unique and often invaluable. Copies of important fossils are put on display in mu-
seums, needed for comparison in research institutions, or used for teaching in schools and
Current developments of digital techniques bring both study and presentation of fossil spec-
imens to a new level (see e. g. CUNNINGHAM et al. 2014). 3D imaging allows to document
the shape, size, optical and – at least partly – structural characters of fossils, making various
measurements and manipulations possible on the screen, i. e. reducing the risk of damaging
the original by touching it, and providing remote collaborators with virtual access to the spec-
imen. Further data can be added, e. g., by incorporation of methods such as CT scanning,
which is a non-destructive means of investigating internal structures of a fossil, or those still
hidden by rock. Digitized fossils can be employed in digital restorations and functional re-
constructions. Thus, digital models offer countless opportunities for research and visualiza-
tion. Furthermore, the increasing number of 3D printing technologies allows for fast and site-
independent reproduction of fossils, plain or even in color and, most importantly, without
affecting the original by the production of silicone or latex peels.
In this contribution, we discuss methodologies for evaluating several 3D imaging and repro-
duction approaches on the example of the type specimen of Madygenerpeton pustulatum
(SCHOCH, VOIGT & BUCHWITZ 2010) (Figure 1). This fossil tetrapod from the Triassic of
One Head, many Approaches 23
Madygen (Kyrgyzstan, Central Asia) is known from a few series of dorsal shields, represent-
ing at least three individuals, and a skull lacking the lower jaw. The skull has been selected
as holotype, i. e., the specimen on the basis of which the species is defined. Madygenerpeton
belongs to the Chroniosuchia, an extinct group of reptiliomorph ‘amphibians’ found in Per-
mian and Triassic deposits of Europe and Asia. Thus, colleagues in various institutions are
interested in the morphology of Madygenerpeton, which can also play a role in classes and
exhibitions as a representative of the reptiliomorph ‘grade’. Further investigation and recon-
struction of the animal, on the other hand, can be facilitated by restoring the digital model
and incorporating missing elements, such as the lower jaw, from closely related forms (see
Fig. 1: The skull of Madygenerpeton pustulatum SCHOCH, VOIGT & BUCHWITZ, 2010 (A)
and a scientific reconstruction of the animal by Frederik Spindler (B)
We created digital 3D models of the Madygenerpeton skull using photogrammetry, struc-
tured light scanning, laser scanning, and computed microtomography. Furthermore, we have
produced printed copies of the skull with various additive manufacturing devices from sev-
eral companies and institutions, based on one of the structured light scans. Both, imaging and
reproduction approaches can be compared with regard to quality, availability, equipment
costs, time consumption and required operator qualification.
2.1 Digitization Techniques
3D digitization means the transformation of physical three-dimensional objects into com-
puter-readable data, such as point clouds, surfaces or volumes with certain accuracy
(HUIJSMANS & JENSE 1993, BARBERO & URETA 2011). Commonly, a cloud of points with x,
y and z coordinates is recorded first and is used in a next step for creating discrete surfaces,
usually by means of triangulation.
We created digital models of the Madygenerpeton skull using the following hard- and soft-
ware solutions (Fig. 2).
24 I. Kogan, M. Rucki, M. Jähne, D. Eger Passos, T. Cvjetkovic and S. Schmidt
Fig. 2: Digitization techniques incorporated in this study. A, photogrammetry with a Fuji-
film X-T2 digital camera; B, structured light scanning with an Artec Space Spider;
C, structured light scanning with a CREAFORM Go!Scan 3D; D, structured light
scanning with an AICON SmartScan industrial camera; E, laser scanning with a
CREAFORM HANDYScan 3D; F, computed microtomography with an YXLON
Photogrammetry (Fig. 2A): some 300 pictures were taken with a Fujifilm X-T2 full-format
system camera with a Fujinon Super EBC XF 10-24 mm 1:4 R OIS lens mounted on a tripod,
which was moved around an illuminated table on which the object was resting. A 3D model
was computed using the commercial software package 3DF Zephyr. The software processes
a point cloud from the pictures (we have obtained around 11.5 million points) and generates
the final photo-realistic textured mesh via surface triangulation.
Handheld structured light scanning (Fig. 2B, C): the object was scanned at various occasions
in different institutions with an Artec Space Spider, an EinScan Pro and a CREAFORM
Go!Scan 3D. 3D models were generated in the respective scanner software.
Industrial structured light scanning (Fig. 2D): the skull was scanned with an AICON Smart-
Scan at the State Archeological Survey of Saxony, Dresden, and with a customized device at
FusionSystems, Chemnitz. 3D models were produced with specially developed software.
Laser scanning (Fig. 2E): the fossil was scanned with the handheld laser scanner CREA-
FORM HandySCAN 3D. A 3D model was obtained using CREAFORM software.
Micro-CT (Fig. 2F): Computed tomography was performed at the Museum für Naturkunde
Berlin using a custom-built YXLON µCT scanner. The reconstructed 3D-gray-value image
from the µCT scanner was then processed with the software package GEODICT (GEODICT
2020). Noise in the image was removed and every voxel of the three-dimensional image was
One Head, many Approaches 25
assigned a distinct material phase: void, sediment, fossil. This 3D-volume model (with sedi-
ment and fossil separated) is the base for further geometrical measurements or visualizations.
In order to create a surface representation of the fossil, the void and sediment phases were
removed and a surface triangulation of the remaining fossil phase was executed (see Fig. 3H).
Fig. 3: Qualitative evaluation of the resolution of digital 3D models. A, holotype of
Madygenerpeton pustulatum in dorsal view, with the parietal region highlighted.
B-H, views of the parietal region under a KEYENCE digital microscope (B) and on
models generated with CREAFORM Go!SCAN 3D (C), Artec Space Spider (D),
CREAFORM HandySCAN 3D (E), photogrammetry (F), AICON SmartScan (G)
and YXLON µCT (H).
While protocols exist for the comparison of 3D scans among each other (see KERSTEN et al.
2016, PETERSON & KRIPPNER 2019), the assessment of 3D scanning accuracy with respect to
the original remains challenging. This is especially true for fragile objects such as fossils,
where contact-based measurement methods cannot be applied. Digital models can be evalu-
ated semi-quantitatively based on the number of points and triangles they consist of (Table
1), and qualitatively by visualization of important structures. Figure 3 shows the triangulated
mesh patterns in such a remarkable region, the area of the parietal opening on the skull roof.
26 I. Kogan, M. Rucki, M. Jähne, D. Eger Passos, T. Cvjetkovic and S. Schmidt
2.2 3D Printing
The term “additive manufacturing”, which is a more precise and inclusive synonym of “3D
printing”, pertains to the fact that, to produce a three-dimensional model, powder or liquid is
consecutively added layer by layer (TOFAIL et al. 2018). Most technological approaches can
be subdivided into powder, extrusion, and UV resin-based concepts. Many of them incorpo-
rate the use of a laser or inkjet-like fluid jetting. We produced and evaluated twelve printed
models of the Madygenerpeton skull (Figure 4) based on a scan of the holotype obtained with
an Artec Space Spider. The technologies applied ranged from desktop extrusion-based FDM
printers (in this case a 750-euro Prusa MK2, print #3), which are easy to use and to maintain,
to 350.000-euro industrial machines (Mimaki, print #12).
Fig. 4: Printed copies of the Madygenerpeton skull, produced with different additive man-
ufacturing methods and materials: 1, 3 – extrusion-based FDM technologies;
2 – Multijet Fusion (polymer powder-based); 4, 6, 7 – 3D printing (powder-based
with inkjet); 5 – UV-resin-Inkjet; 8, 9 – Polyjet; 10, 11 – ColorJet printing; 12 –
UV-curable inkjet printing
2.3 Metrological Evaluation
The holotype of Madygenerpeton and the printed models have been measured with a Mi-
tutoyo Coordinate Measuring Machine (CMM) at Mitutoyo Polska, Wrocław. The CMM was
CRYSTA-Apex S 9166 of 900 × 1600 × 600 mm range and maximum permissible error
MPEE= ±(1.7+3L/1000) μm. The non-contact line laser probe SurfaceMeasure 606 was ap-
plied for surface scanning. Its scanning error was 12 μm [1σ/ sphere fit]. For each point, the
distance between the respective points of the original fossil and the printed model was calcu-
lated and represented with a color in the visualization. As can be seen in Figure 5, presenting
the comparison of the original fossil with printed model #6, these distances ranged from –1
to 1 mm, marked in the graph with blue and red, respectively. Example of a point marked
green (Fig. 5B) specifies this distance as 3D = –0.3627 mm and provides also respective
distances along each axis dX, dY and dZ.
One Head, many Approaches 27
Fig. 5: Color map of deviations between the original fossil surface and the printed model
#6 (A) and detailed example of its one point marked green (B)
Measurements have also been attempted with a more accurate device. The device was CMM
STRATO-Apex 574 with measuring range 500 × 700 × 400 mm, maximum permissible error
= 0.7+2.5L/1000 μm and 5 μm scanning error for roundness. It was equipped with a
non-contact line laser probe SurfaceMeasure 201FS with a 1.8 μm scanning error. These
attempts, however, proved unsuccessful, although the obtained cloud of points looked very
promising, but the triangulated model showed numerous discontinuities and was very far
from the original image (Fig. 6).
Triangulated model obtained from the
cloud of points from CMM STRATO-
To obtain an accurate measurement surface from the cloud of points, it would be required to
perform additional digital operations and perhaps to use other software. However, this was
found unnecessary because of satisfactory accuracy of the measurements performed with
CRYSTA-Apex S 9166 CMM and SurfaceMeasure 606 probe.
3 Preliminary Results
3.1 Digitization Techniques
Preliminary evaluation of several digitization techniques on the example of the holotype of
Madygenerpeton pustulatum shows that, along with industrial 3D scanners, satisfactory re-
sults can be obtained using some handheld light-optical devices, photogrammetric ap-
proaches, laser scanners and computed microtomography. The Artec Spider structured-light
scanner delivered a 3D model that could be used for fabrication of highly accurate 3D prints
28 I. Kogan, M. Rucki, M. Jähne, D. Eger Passos, T. Cvjetkovic and S. Schmidt
(see below). Further advantages of Artec Spider are its accessibility (purchase costs of about
20.000 euro), user-friendliness and the low amount of necessary post-processing. Other
handheld light-optical scanners required more complex manipulation, longer computation
times, and produced models of lower quality (Fig. 3, Tab. 1). In the frame of our study, pho-
togrammetry also was found inferior to 3D scanning with respect to preparation and espe-
cially post-processing time, but clearly delivered the most photo-realistic image. The most
expensive method, µCT, is unrivalled in revealing internal structures of a fossil or those cov-
ered by sediment, and, thus, provides the highest scientific benefit (Fig. 7).
Table 1: Comparison of the applied digitization devices and 3D models derived from
3.000 2 h >> 5 h 299.910 599.453
EinScan Pro struct. light 8.000 3 h > 2 h 6.257.953 12.515.902
Artec Spider struct. light 20.000 2.5 h < 1 h 277.806 555.608
struct. light 30.000 < 1 h < 0.5 h 58.398 116.457
AICON struct. light 100.000 ca. 1 h ca. 1 h 6.957268 13.913.354
laser 40.000 0.5 h < 0.5 h 1.563.656 3.117.642
YXLON µCT 1.200.000 2 h > 4 h 8.071.807 16.181.930
Fig. 7: 3D model of the skull of Madygenerpeton pustulatum with sediment removed, in
dorsal (A) and ventral (B) view. Boxes in B highlight series of palatal teeth, entirely
covered by sediment and therefore unknown in Madygenerpeton prior to µCT scan-
3.2 3D Printing
Models printed with different technologies out of different materials attained different accu-
racy of fossil reproduction. The smallest deviation range was detected in the Polyjet-printed
model #9 presented in Figure 8. It can be seen from the comparison pie that very few points
lay above or below the area between –0.4 and 0.4 mm compared to the original fossil surface,
the mean square root of deviation is σ = 0.2546 mm.
One Head, many Approaches 29
Fig. 8: Deviations between the original fossil surface and the printed model #9
Fig. 9: Deviations between the original fossil surface and the printed model #7
The least reproduction accuracy was found for the model #7 shown in Figure 9, produced
with PowderBed-Inkjet technology from apricot kernel powder. Large percentage of points
above 0.4 mm distributed throughout the surface and many points below –0.4 mm in the
central area of the surface demonstrate large deformation of the model #7. The mean square
root of deviation is σ = 0.5058 mm, twice as large as in case of the model #9.
Remarkably high accuracy was also attained by model #3, generated with a low-cost FDM
printer from Josef Prusa set at highest resolution (Fig. 10). This shows the potential of home
3D printers for high-quality replication.
30 I. Kogan, M. Rucki, M. Jähne, D. Eger Passos, T. Cvjetkovic and S. Schmidt
Deviations between the origi-
nal fossil and printed model #3.
ote that the orbits, here
marked in blue, have been
milled manually for the pur-
pose of an exhibition.
4 Discussion and Conclusions
Digitization results presented in this contribution must be considered preliminary due to time,
software and hardware limitations. Several approaches, such as photogrammetry and µCT,
required the use of additional software, which is not freely available. Furthermore, depending
on the number of scanning frames, the mesh size of the triangulated model and the program
version used, we encountered computation difficulties. Another factor that has not been sys-
tematically addressed is the operator qualification or experience. It is clear, however, that
little training is needed to correctly use a handheld Artec scanner, while a µCT machine
should only be operated by well-prepared and experienced staff. Nonetheless, low-cost solu-
tions such as EinScan Pro or photogrammetry also require much user experience in order to
deliver useful results.
Evaluating the accuracy of 3D scans via the comparison of 3D prints with the original object
has proven an easy and seemingly robust method, which can be explored further by printing
and metrologically comparing 3D models from different digitization devices. However, some
uncertainty persists with respect to possible deviations between the digital model and the
Not all digitization and replication approaches produced results that could fully be integrated
in the study. For instance, the photogrammetric model considered here only covers the dorsal
side of the Madygenerpeton skull. The printed copy #5 could not be measured with the CMM
because of its translucent surface.
We conclude that, although the trends revealed in this study on 3D digitization and replication
seem reliable, more research is needed to prove their significance and reproducibility.
We are indebted to Birgit Gaitzsch (TU Bergakademie Freiberg) for providing access to the
holotype of Madygenerpeton pustulatum, to Michael Buchwitz (Museum für Naturkunde
Magdeburg) for advice on chroniosuchids, to Kristin Mahlow (Museum für Naturkunde Ber-
One Head, many Approaches 31
lin), Henrik Ahlers (SLUB Dresden) and Thomas Reuter (Sächsisches Landesamt für Ar-
chäologie Dresden) for help with digitization, to Christina Burkhardt, Henning Zeidler (TU
Bergakademie Freiberg) and the team of the SLUB Makerspace for production of printed
models, and to the staff of Mitutoyo Polska for measurements. This contribution presents
results achieved within the ESF-funded young researcher group “G.O.D.S.” (Geoscientific
Objects Digitization Standards), initiated by Gerhard Heide at the TU Bergakademie Frei-
berg. Work of first author was performed according to the Russian Government Program of
Competitive Growth of Kazan Federal University.
Barbero, B. R. & Ureta, E. S. (2011): Comparative study of different digitization techniques
and their accuracy. Computer-Aided Design, 43 (2), pp. 188-206.
Cunningham, J. A., Rahman, I. A., Lautenschlager, S., Rayfield, E. J. & Donoghue, P. C. J.
(2014): A virtual world of paleontology. Trends in Ecology and Evolution, 29 (6), pp.
GeoDict, Release 2020, Math2Market GmbH, Germany. https://www.geodict.com.
Huijsmans, D. P. & Jense, G. J. (1993): Recent Advances in 3D Display. Advances in Elec-
tronics and Electron Physics, 85, p. 77-229.
Kersten, T. P., Przybilla, H.-J. & Lindstaedt, M. (2016): Investigations of the geometrical
accuracy of handheld 3D scanning systems. Photogrammetrie – Fernerkundung – Geoin-
formation, 5-6/2016, pp. 271-283.
Lautenschlager, S. (2016): Reconstructing the past: methods and techniques for the digital
restoration of fossils. R. Soc. open sci., 3, 160342.
Peterson, J. E. & Krippner, M. L. (2019): Comparisons of fidelity in the digitization and 3D
printing of vertebrate fossils. Journal of Paleontological Techniques, 22, pp. 1-9.
Schoch, R. R., Voigt. S. & Buchwitz, M. (2010): A chroniosuchid from the Triassic of Kyr-
gyzstan and analysis of chroniosuchian relationships. Zoological Journal of the Linnean
Society, 160, p. 515-530.
Tofail, S. A. M., Koumoulos, E. P., Bandyopadhyay, A., Bose, S., O’Donoghue, L. & Char-
itidis, C. (2018): Additive manufacturing: scientific and technological challenges, market
uptake and opportunities. Materials Today, 21(1), p. 22-37.