Contents lists available at ScienceDirect
Journal of Archaeological Science: Reports
journal homepage: www.elsevier.com/locate/jasrep
A white-box framework to oversee archaeological virtual reconstructions in
space and time: Methods and tools
Emanuel Demetrescu*
, Bruno Fanini
CNR-ITABC Istituto per le Tecnologie Applicate ai Beni Culturali Via Salaria, km 29.3 - Monterotondo 00015, Roma, Italy
A R T I C L E I N F O
Keywords:
3D semantic annotation
3D virtual reconstruction
Extended Matrix
Virtual Stratigraphic Unit
Scenegraph
A B S T R A C T
The goal of this paper is to present original methods and visual tools able to formally document the scientific
processes behind an archaeological virtual reconstruction, namely a new version of the Extended Matrix (EM
1.1) and the Extended Matrix Framework (EMF 1.1). The proposed approach aims to improve the EM as well as
methods and tools for 3D query, visualization, and inspection of extended matrices in order to solve current
bottlenecks and issues with the integration of 3D virtual environments and rich semantic descriptions (EMF). A
real case scenario is provided to present the steps involved in a reconstruction project using EM/EMF: the Great
Temple of the ancient Roman town Colonia Dacica Sarmizegetusa.
1. Introduction
The goal of this paper is to present original methods and visual tools
which can formally document the scientific processes behind an archaeological
virtual reconstruction, namely the Extended Matrix version 1.1
(EM 1.1). The “scientific process behind an archaeological reconstruction”
includes not only the specialists' hypotheses regarding a specific
context in order to reconstruct its “original aspect” in a given historical
time period, but also the collection and the use of the “sources”. The
sources comprise all the physical evidence (such as discoveries made
during excavation or pieces from a museum's collection), historical
assumptions (i.e. in Roman times outside the city wall and along the
main road there was a necropolis), and documents (drawings or writings)
from the past that testify to the lost aspects of the archaeological
site as well as its 3D survey and stratigraphic reading.
EM, as defined in its 1.0 version (Demetrescu, 2015), is a formal
language specifically designed for the reconstruction of lost contexts1
and operates with the same tools already in use in the archaeological
domain in order to manage time sequences (matrix of Harris) and data
granularity (stratigraphy). The version 1.1 of the Extended Matrix adds
a complete support for 3D representation of extended matrices and
scientific publication of the whole dataset behind a virtual reconstruction
hypothesis. The EM 1.1 is focused on solving bottlenecks
and issues that are currently present in the integration of 3D virtual
environments and rich semantic descriptions (see infra sez. 3). Examples
from a real case scenario are provided to present the steps involved in a
reconstruction project using EM: the Great Temple of the ancient
Roman town Colonia Dacica Sarmizegetusa.
2. Theoretical domain of the research
Arguments like data provenance, uncertainty management, and
data modelling are core topics when it comes to source based models2
(reconstruction of lost contexts). The intricacy of the sources and reasoning
behind a reconstruction hypothesis may blur the scientific
communication of the archaeological process. This situation needs an
efficient “white-box” system that permits easy data management, rich
data ingestion, and expert data retrieval through visual tools.
The integration of the EM with digital tools for 3D representation of
virtual reconstructions and visual inspection of extended matrices is
identified with the expression Extended Matrix Framework (EMF). The
EM is the theoretical and methodological background, usable even
outside a digital environment (an EM can be sketched out on a sheet of
paper with a pencil as well as with cutting-edge software) while the
EMF includes digital solutions and software platforms (which can
substantially change in the future if relevant technological innovations
occur). EM is about scientific-driven content creation, EMF is about
technological-driven solutions. Every version of the EM drives its own
EMF digital framework identified by the same number (at present EM
1.1 has its EMF 1.1).
http://dx.doi.org/10.1016/j.jasrep.2017.06.034
Received 2 March 2017; Received in revised form 17 June 2017; Accepted 18 June 2017
*
Corresponding author.
E-mail addresses: emanuel.demetrescu@itabc.cnr.it (E. Demetrescu), bruno.fanini@itabc.cnr.it (B. Fanini).
1
By lost context we mean a context that is no longer existent and/or has been completely or partially “lost” over time due to various causes (war, neglect, environmental factors, etc.).
2
A 3D reconstruction can be thought of as a complex “source driven 3D model” because it is based on several “sources” (archaeological records, images, comparatives, etc.) combined
together according to an analytic and comparative methodology (Demetrescu, 2015, p.43).
Journal of Archaeological Science: Reports 14 (2017) 500–514
Available online 27 June 2017
2352-409X/ © 2017 Elsevier Ltd. All rights reserved.
MARK
The proposed EM-EMF 1.1 includes the following aspects:
1. A coherent language that makes it possible for various professionals
to work together, following a smooth validation work-flow and a
collaborative pipeline. EM allows archaeologists, IT scientists, and
computer graphic experts to share a common language which ensures
that the scientific reasoning behind the reconstruction hypothesis remains
consistent and transparent (see infra sez. 4);
2. Methods and tools to populate a graph database (built on the
Extended Matrix 1.1) with rich semantics intended to visually describe
the elements involved in the reconstruction (sources, concepts,
comparisons, see infra sez. 4 and sez. 4.3);
3. Foundations for the crafting of 3D visual inspection tools in order to
query and validate Extended Matrices at runtime, addressing (a)
temporal dimension, (b) automatic graph extraction from EM definition,
and (c) visual presentation of meta-data elements (see infra
sez. 5).
EMF is a multidisciplinary framework that operates both within the
“historical“ and “mathematical” domains (see the classification of
Nicholaus Steno, Ascani et al., 2002). Through the historical language
we can turn the reconstructive record3
into an EM data model (the EM
describes the elements and their relationships). On the other hand,
through the mathematical language it is possible to consistently inspect
data and represent it dynamically in a virtual world, enabling the exploration
and experimentation of the reconstruction hypothesis within
a 1:1 scale virtual space “shared” with other researchers.
The Extended Matrix is a semantic graph that leads to a schema-less
data model: the reconstructed objects and their descriptive elements are
heterogeneously fitted into space and time, in a way that better suits the
incompleteness of the historical record. The descriptive elements (EM
nodes, see infra sez. 4.2.1) are used as a modular grammar to compose
the final description of the reconstruction process (data-driven reconstruction).
Let's look at an example: I could describe a USV using
just the property “material” (i.e. a wooden lintel) because it is the only
reconstructive value I am confident in. Meanwhile, in the case of other
USVs I may declare more properties, each of which can be validated by
different sources. There is no predetermined schema: each USV has its
own unique node tree (whiting a common EM data structure) which
describes and validates the USV itself (see infra paradata nodes at Fig. 5
(g), (h), (i), and (j)).
In the following sections, after a review of the state of the art regarding
the topics mentioned above (see infra sez. 3), the theoretical
and methodological background of the EM 1.1 as well as of the EMF 1.1
which include tools for 3D visualization and inspection will be presented
(see sez. 4 and 5). In order to clarify them, examples from a real
case scenario (the Roman Great Temple at Colonia Dacica
Sarmizegetusa) will be provided in order to describe the steps involved
in a virtual reconstruction project: from data collection (sez. 4.1) to EM
creation (sez. 4.2), 3D modelling(sez. 4.3), and 4D data exploration and
visualization (sez. 5).
3. Related work
3.1. Managing data provenance in archaeological virtual reconstruction
In the last few years an increasing number of experts have been
paying attention to data provenance in (3D) virtual reconstruction (see
recent introduction of classification proposals: Münster et al., 2016a,
Münster et al., 2016b, Apollonio, 2016, Pfarr-Harfst, 2016). A common
approach used to represent the thought processes behind the creation of
a (3D) reconstruction is to annotate them using meta-data descriptions.
For a critical review about data provenance strategies and data
granularity in archaeological virtual reconstruction see Demetrescu,
2015, pp. 43–44.
At present, there are several semantic tools that describe existent
objects of cultural heritage, each specialized in specific domains
(monuments, objects, discoveries made during excavation, bibliography,
etc.) and organized according to conceptual reference models
(CIDOC-CRM, CHARM).
CIDOC-CRM has introduced a high degree of stability to meta-data
enrichment of tangible objects (mainly collections) thanks to the use of
a common standard. However, for the purpose of standardization, this
ontology was first developed with specific objectives and with specific
restrictions: “objects in museums, libraries, and archives”(Doerr, 2003,
p. 84).
The process of metadata ingestion in the field of the reconstruction
of lost contexts has a different workflow from the 3D survey and digitalization
of real objects (see Fig. 1). It follows a circular process and
results in an “open” output: the virtual reconstruction never becomes a
“definitive” output because it can be modified at any time in the future
if new sources become available.
At the same time the qualification of the provenance of the data
follows a completely different path. When it comes to the digitalization
of real elements, the “accuracy” can be expressed in quantitative units of
measure (i.e. 2 mm) while in the case of virtual reconstruction the accuracy
is a qualitative blending of different sources (it cannot be expressed
in discrete units of measure). The focus of the semantics that
describe virtual reconstruction processes is the network of the relationships
between the sources involved (see infra sez. 4).
Despite the fact that CIDOC-CRM is specialized in handling the
properties of tangible objects, in recent years some additions have been
proposed which would extend its core-elements to include more abstract
concepts and relationships.
CRM-sci is intended to support scientific observation and description
of physical phenomena (Doerr et al., 2014). CRM-BA is intended to
describe the elements and relationships involved in building archaeology
(Ronzino et al., 2015). CRM-dig is a generic digital provenance
model for scientific observation and can be used to create data paths to
information provenance of digital elements from real world objects
(Doerr and Theodoridou, 2011) and it has been used also to annotate
reasonings behind a reconstruction (Bruseker et al., 2015).
In recent years, other tools and standards have been proposed with
the specific purpose to annotate virtual reconstruction processes using
different paradigms (stratigraphic approach combined with visual tools
like the Extended Matrix: Demetrescu, 2015, Demetrescu et al., 2016),
conceptual models (CHARM: Gonzalez-Perez et al., 2012, Apollonio and
Giovannini, 2015) or customized formal languages (CHML: Hauck and
Kuroczyński, 2014, Kuroczyński et al., 2016). In these cases, CIDOCCRM
has not been used as a semantic reference.
In 2012 CHARM was introduced, a conceptual reference model
based on a visual representation: the ConML language. This feature sets
CHARM apart from CIDOC-CRM.
“[...] CHARM aims to cover [...] cultural heritage [...] at a high level of
abstraction. In contrast to CIDOC CRM, CHARM is much wider, since it
does not focus only on the curated knowledge of museums but on cultural
heritage in general. At the same time, it is much shallower, because of its
high level of abstraction; CHARM has been designed under the assumption
that extension mechanisms need to be applied before it can be
used at all [...], whereas CIDOC-CRM attempts to be an off-the-shelf
solution.” (Gonzalez-Perez et al., 2012, p.191).
In CHARM, concepts like the Stratigraphic Unit (SU) and its relationships
are already provided for. CHARM offers a higher level of
abstraction and is modular, which means that each element can be
easily modified for other purposes. On the other hand, the Cultural
Heritage Markup Language has a reconstructive point of view that
provides a visual language with which to describe a virtual reconstruction
(Kuroczyński et al., 2016 and Kuroczyński et al., 2014)3
See definition of reconstructive record at the end of sez. 4.
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
501
inspired by the Unified Modeling Language (UML) that is object-oriented
and focused on the generic description of processes and their
evolution in time (Rumbaugh et al., 2004).
Some research considers the organization of the sources of the reconstruction
and the annotation of the steps in the creation of the reconstructive
3D asset (Baldwin and Flaten, 2011, Münster, 2013, PfarrHarfst
and Grellert, 2016) to be a necessary starting point for data
transparency. To this end, a wiki system which collects data about
virtual reconstruction has recently been proposed (Münster and
Niebling, 2016).
3.2. Linking 3D models and data
The 3D annotation has become, over the last decade, a major topic
when it comes to accurately documenting an entity of cultural heritage.
Common approaches within the interactive graphics domain applied to
cultural heritage include different processes that define query-able descriptors
in the form of basic 3D shapes (Serna et al., 2012), semantic
structures produced by 3D segmentation of digitized models (Apollonio
et al., 2012), and tools offering multiple annotation modes (i.e. point,
surface, and frustum, see Shi, 2016). In terms of interactive on-line
fruition by means of modern WebGL capabilities, previous research on
interactive queries has focused on 3D query-able geometrical shapes
integrated in web-based viewers (Auer et al., 2014, Potenziani et al.,
2015).
3.3. Scientific publication of virtual reconstructions
Despite the fact that a thorough publication of the data related to
virtual reconstruction is considered a primary issue in the field, as of yet
there is no common standard to accomplish this task. An important
aspect is the difficulty in making not only the sources used available,
but also the reasoning behind a reconstruction hypothesis (Bruseker
et al., 2015). Web-based tools oriented to track the processes behind a
virtual reconstruction (Bruschke and Wacker, 2016) stress the importance
of connecting the source with the hypothesis (Pfarr-Harfst and
Grellert, 2016), while others are committed to showing reconstructions
on-line (Pompei: Dell’Unto et al., 2013, Çatalhöyük: Forte et al., 2012,
MayaArch3D: von Schwerin et al., 2016) through a 3D GUI focused on a
hierarchical organization of the digital collections. In some cases the
researchers combine the release of the project's stratigraphic database
with the synthetic publication of the excavation and include the reconstruction
as part of the archaeological investigation using the Extended
Matrix (The Swedish Pompei Project: Landeschi et al., 2016,
Demetrescu et al., 2016). In some cases the GUI is also focused on the
study of user perception (The Gabii Project: Opitz and Johnson, 2016).
For a broader state of the art about the projects focused on GUIs for the
exploration of 3D models of archaeological excavations and related
data see Opitz and Johnson, 2016, p.7.
4. Definition of norms, tools and practical aspect of EM 1.1
workflow
In this section we will present the guidelines (norms, terms,
methods) for the EM 1.1 work-flow. At the end of the section, the Great
Temple of Colonia Dacica Sarmizegetusa (Romania) will be presented
in order to provide an example of the practical actions and tools used in
“standard scenario”.
The guidelines include the steps in the documentation of a virtual
reconstruction (see upcoming numbered statements in Fig. 2):
1. Data collection: 3D survey of the remains and of the special findings
(i.e. non in situ blocks of stone from the architectural apparatus),
stratigraphic reading (4D analysis of the palimpsest), creation of a
Dossier Comparatif (comparative study);
2. EM model creation to organize data within the Extended Matrix:
writing down of the Extended Matrix and the Report of Virtual
Activities;
3. Creation of 3D models: The survey creates a 3D snapshot (see infra
6) of a monument at a specific moment in time (i.e. 2017) that represents
a palimpsest of different phases of life (i.e. prior to construction,
construction, destruction, modern restoration). It has to
be segmented in order to provide different geometrical bases for the
subsequent 3D source based modeling. The EM must be divided into
chronological horizons and for each one, a reconstructive model is
proposed, filling the gaps highlighted in the 3D snapshots (negative
stratigraphic units).
We do not have to think of these steps as separate phases of a reconstruction
project: during data collection we can start populating the
EM with interpretative elements as well as sketch out some visual representations
of a given epoch. This synchronous work-flow enables a
cross-fertilization of the relationships between the archaeological elements,
the external sources, and the hypotheses. We can call the information
stored during the above-mentioned steps an archaeological
reconstructive record. The documentation of a reconstructive record that is
coherent and as complete as possible is the primary scope of the EM
work-flow.
4.1. Data collection
Data collection can include diversified methodologies and technical
solutions. Different contexts require different approaches. Some scenarios
cannot involve a 3D survey due to disparate limitations (physical,
legal, etc.) but can make use of legacy data (drawings or blueprints).
Furthermore, when the reconstruction project regards a
completely lost context, all the data collection is focused on the sources
organized within the Dossier Comparatif (DosCo) (see infra 4.1.3).
4.1.1. 3D Survey of the context
An archaeological site is a palimpsest of different epochs in which
Fig. 1. Metadata from a workflow point of view: reality
based and source based scenarios of metadata creation.
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
502
each of these epochs must be purged of non-coeval elements in order to
highlight the preserved ones and the overall shape of each chronological
phase. Furthermore, a clear distinction between the different
stratigraphies (grouped by epoch of belonging) is crucial to proposing a
reconstruction hypothesis. In Fig. 3 there is an example of digital 3D
geometry acquisition of the Great Temple in Sarmizegetusa.
Alongside the in situ elements, the 3D acquisition could include the
survey of all the non in situ objects. The 3D measurements result in
metrically accurate 3D models or digital replicas with colour information.
In the example of the Great Temple in Fig. 4, every architectonic
fragment has been formalized in a high resolution blueprint. In this
case, each element is represented through different orthometric views
(front, back, right, left, top, bottom), its measurements are provided
(depth, width, height), cross sections are cut, and, finally, joint-areas
are highlighted in order to check its topological compatibility with the
remains of other stone blocks.
4.1.2. 4D reading: the stratigraphic sequence and the Matrix of Harris
During (or after) the 3D survey, a stratigraphic reading must be
performed in order to highlight the temporal sequence (4th dimension)
of the actions. In the example of the Great Temple in Fig. 13, the Matrix
of Harris (green areas) visually summarizes the stratigraphic elements
involved and the epochs which they belong to. In this case the stratigraphic
reading has been performed on the field and includes the
building stratigraphy.
4.1.3. Dossier Comparatif
All the sources and the comparisons with other contexts are stored
in the Dossier Comparativ (comparative study Gros, 1995, p.322). The
Dossier Comparatif (DosCo) is a collection of documents which follows a
specific nomenclature: a composite known as “D.”(Document), plus an
increasing number (i.e. D.01). The number is set according to the cronological
sequence of data ingestion. All the documents are linked inside
the EM and used to validate the reconstruction hypothesis. These
are represented inside the EM through the source node (Demetrescu,
2015, pp. 48–50).
4.2. Creation of the Extended Matrix and the Report of Virtual Activities
In archaeological practice, a virtual reconstruction is usually created
at the end of an excavation project or 3D survey and is not part of the
research itself. The 3D model, in this sense, is intended as a posthumous
work which synthesizes different hypotheses made during the investigations
(Medri, 2003). Generally these are not formally annotated.
In some cases, experts write down intermediate reports to fix certain
general ideas, but the collection of the archaeological record lacks a
precise way to store the rich connections that are recognized between
pieces of evidence during the fieldwork. The EM is committed to fixing
this issue.
Virtual reconstruction should be part of the archaeological investigation
right from the earliest stages of research. Even if the final
results are not to be shared until the end of a survey/excavation, the
reconstruction hypothesis must be formally “documented” along with
the archaeological record which it is derived from. “Early collection”
simplifies the management of the reconstructive record and “drives” the
subsequent source-based modelling. This approach makes it possible to
publish a thorough account of the scientific process, highlighting the
connection between the archaeological and the reconstructive records
and enabling other researchers to actively check, modify, and consciously
reuse the virtual reconstruction. The Extended Matrix was used
during the survey of the Great Temple in order to manage the sources,
acting as a “mind map” of the researcher's intuition and of the connections
discovered in case a new document is added to the research.
Besides the Extended Matrix, a more discursive document, the Report of
Virtual Activities, provides a textual version of the EM for quick sharing
of the reconstruction hypothesis (see infra sez. 4.2.2).
Fig.2.Productionwork-flow,fromdatacollectionto3Dvisualization.
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
503
4.2.1. The grammar of the Extended Matrix
EM uses two sets of standardized nodes4
: USV and validation nodes
(see Fig. 5). The USV nodes represent virtual stratigraphic units (according
to a specific typology) while the validation nodes express the
reconstruction process behind the USV. These nodes are connected to
each other by arcs in the same way as with the archaeological Matrix of
Harris.
Validation nodes (or paradata nodes, see Fig. 5 (g), (h), (i), and (j))
have unique names (as well as the USVs) in order to be correctly referenced.
They adhere a name convention model (see Fig. 2): extractor
nodes are composed of a “#” plus a sequence of numbers (i.e. the first
extractor of an EM will be #01). Combiner and document nodes use the
“$” and “D.”.
The EM shows the epochs of the site. Each epoch is divided into two
rows (fig. 13): one row for the physical remains (SUs) and another row
for the reconstruction hypothesis (USVs). An EM including only the
physical remains is equivalent to a standard Matrix of Harris.
Fig. 3. 3D model of the Great Temple (photogrammetric survey).
Fig. 4. Example of a blueprint with a fragment of a column (Special Finding T01).
4
For full details about EM nodes and their use, see Demetrescu, 2015, pp. 47–50.
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
504
4.2.2. Report of Virtual Activities
As happens with real stratigraphic units, Virtual Reconstruction
Units (USV) are combined in Virtual Activities (VAct. see Fig. 11) using
rectangular shapes in the EM canvas. VActs are described in the Report
of Virtual Activities (see Demetrescu, 2015, p. 51), a textual report that
acts as an intermediate output and is useful for quickly sharing a reconstruction
hypothesis. A Report of Activities is written down for each
reconstructed epoch (see Fig. 2). In the following text box there is an
extract of the VAct. 3 from the Report of Activities of the Great Temple.
Box 1 Virtual Activity 3: lintel over the columns. In this
reconstruction it is presumed that over the colonnade USV104
there was a wooden lintel with stuccoes USV106. According to
Vitruvius (Vitr. III,2), the diastil module manner is fragile
when we come to the lintel and mentions collapses which
happened in the past. For that reason, for this kind of module,
he suggests a wooden lintel (with stuccoes for the decoration).
In other words, in this kind of module we can expect both a
wooden and a stone lintel. The lack of stone blocks during the
recent excavations of the lintel suggests however that it may
be wooden. The overall width of the lintel is normally 1/4 or
1/5 of the whole length of the column.
4.3. Creation of 3D models (source-based modelling)
In this section the 3D modelling steps, starting from the sources
organized in the EM (source-based modelling), are described. These
have different levels of representation, from small details to a broad
perspective:
• Digital restoration of digital replicas (see Fig. 8);
• Digital anastylosis of the elements in order to restore their original
position and spatial relationships (see Fig. 9);
• Virtual Reconstruction of a given epoch starting from all the sources
available and the anastyloses performed (see Fig. 10). It is important
to take into account that from a reality-based model it is possible to
SU 9
(a) (b) (c)
SF 1
(d) (e)
USV
(f)
(g) (h) (i) (j)
Fig. 5. Examples of symbols used in the Extended Matrix: (a) Stratigraphic Unit or “SU”; (b) Virtual Stratigraphic Unit or “USV/S”, related to a structural gap; (c) USV/N related to a non
structural gap; (d) SF (Special Find) not in situ; (e) USV serving as a representation of an SF not in situ; (f) Seriation node; (g) Extractor node, capable of extracting specific information
from the sources and transforming them into properties of the USV; (h) Combination node, useful in combining two or more extractor nodes; (i) Property node, validates the USV it is
connected to; and (j) source useful for the reconstruction (text, image, etc.).
A
B
C
Fig. 6. Relation scheme between representation models, proxy models, and Extended
Matrix (A1-A3, B1-B3); and an example of query on the 3D model selecting the corresponding
USV in the EM (C1-C3).
Fig. 7. Section N-S of the Great Temple's proxy - II A.D. epoch (see its EM at Fig. 13).
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
505
make several reconstruction hypotheses: ideally at least one for each
period formalized in the Matrix of Harris (see Fig. 12);
Despite the fact that it does not contain geometry information, for
the sake of clarity, we summarize here the next level of data representation
of this series:
• Extended Matrix is the highest level of synthesis and simultaneously
represents the site in all its epochs and hypotheses, highlighting the
elements involved in the reconstruction and their relationships: it is
a semantic of the virtual reconstruction (see Fig. 11) and is based on
the relative sequence of actions (stratigraphy) or deduced from the
sources.
4.3.1. The proxy-representation model approach
The creation of 3D content follows two levels of abstraction: proxy
models (see Fig. 7) which are simplified representations of the reconstruction
through basic geometrical shapes (cylinders, boxes,
spheres, etc.) and representation models (see Fig. 10) which are focused
on fine geometries, colour, and material simulations resulting in the
final, aesthetic depiction of the reconstruction hypothesis. The Proxy
level makes it possible to highlight portions of the representation model
and to interact with it (see Fig. 6 and sez. 5). Furthermore, it also represents
the first draft model of the reconstruction and is useful, along
with the Report of Virtual Activities and the EM, for the sharing of intermediate
results with colleagues and other experts and for providing
feedback.
The Extended Matrix contains the elements of all the 3D models (a
model for every reconstructed epoch). Each USV node of the EM is
linked, on a one-to-one basis, to a specific proxy geometry in the 3D
model. A common issue in 3D data granularity is the segmentation of
the meshes. A model created using a computer graphics approach (like
the representation model) may indeed be quite different from a realitybased
model, resulting in an even more challenging segmentation
process. A very popular solution in source-based modelling is “polygonal
modelling with control point” in which the resolution is usually
parametric and controlled by “modifiers”(mirror, subsurf, bevel, etc.).
In general, the computer graphic approach requires an accurate topology
in order to guarantee render time (in the case of a computer
animation) or FPS (in the case of a real-time application) remaining in a
“green zone” of performance. In this type of situation a geometric
segmentation of the model may compromise its overall “usability”. For
this reason, the EM approach adheres to the solutions proposed by the
3D-COFORM project (Serna et al., 2012) for reality-based models. In
this case a proxy 3D model is used to semantically enrich the representation
mesh. In Fig. 6, for instance, the extended matrix (A1) of a
temple has a proxy geometry (A2) which highlights and semantically
enriches the 3D representation model (A3). Each USV in the EM (B1)
may be individually selected, with the corresponding 3D proxy model
(B2) highlighting the details in the representation model (B3) with a
different colour and providing a conceptual visualization. At the same
time, while exploring the 3D model (C1), the EM (C3) can be retrieved
through a pop-up menu (C2).
4.3.2. Graphical documentation for the reconstruction hypothesis
As said before, the reconstruction hypothesis of a context has two
depictions: a proxy model and a representation model. These enhance
communication of research results and make it possible to explore and
experience it in a virtual space (this will be discussed in detail in
Section 5). The EM approach however is not focused only on digital
media but it pertains, first of all, to the archaeological documentation
for a site's interpretation and scientific publication. Each stratigraphic
unit - virtual or real - has a proxy representation. Making sections and
plans out of proxy models is possible in order to provide technical
documentation along with the description of virtual activities: in Fig. 7
the numbers of the SU/USV are labels of the proxy elements while
colours follow a standardized chart (see Demetrescu et al., 2016, p. 55)
based on the typology of SU/USV (remains, USV/n, USV/s, anastylosis,
etc.). Extended Matrix, Report of Virtual Activities, and graphical
documentation published together allow for a coherent formal representation
of the reconstruction hypothesis.
Fig. 8. Digital restoration (USV 101) of a digital replica
(SF 01). On the right the EM representation.
EM
Fig. 9. Digital anastylosis (USV 125) of the special findings (T35, T28, T04, T27, T32,
T15) in order to restore their original position and the spatial relationships. In the lower
part of the figure it is its EM representation.
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
506
4.4. The 3D-as-a-container metaphor and its limits: the multidimensional
approach of the EM
A very common approach when using 3D technology for scientific
applications is to “link data to 3D elements” In this case the role of the
3D space is to “contain” and manage all information. It represents the
fulcrum around which all the sources are structured. This approach
derives from the well known metaphor of space as a container like in
the case of BIM (Building Information System) or GIS (Geographical
Information System). The idea behind the Extended Matrix is to reverse
the 3D-as-a-container perspective switching the fulcrum from space to
time making a multidimensional (3D + 4thD + different self-excluding
hypotheses) approach possible, in which the container of all the reconstructive
processes is not 3D space but a semantic network (the
Extended Matrix). In this way the focus is no longer on the 3D model itself
but on all the possible models derived through data and reasoning. Another
result is that the creation of a 3D digital model is no longer the first
attempt at virtual reconstruction since the majority of the work can be
done (or at least organized) in the mind of the specialist in charge of the
reconstruction with the help of the EM. During the study, the
Fig. 10. Virtual Reconstruction of the II A.D. epoch of the Great Temple starting from all the sources available and performed anastyloses (see Fig. 9 ).
Fig. 11. Extended Matrix represents the Great Temple in
all its epochs and hypotheses, making explicit all the elements
involved in the reconstruction and their relationships.
In red, the Virtual Activities. For a detail (with the
VAct. 1), see Fig. 13.
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
507
researcher's eye reads and interprets the sources, finds similarities, and
performs a task it is very good at: making connections between pieces of
data. The “mind map” of the researcher is “sketched out” through the
EM language: it is the fulcrum of the reconstruction and makes it possible
to organize all the sources in the 4th dimension (y axis of the
oriented graph). The EM can lead to several reconstruction hypotheses:
one model for each chronological horizon to be represented or, in the
case of persisting doubts and equivalent hypotheses, a chronological
horizon can spread out in different models, one for each, self-excluding,
hypothesis. If the EM contains all the reconstruction information, the
resulting 3D model follows a “data-driven” work-flow: changing
sources and reasonings in the EM will result in new proxy and representation
models.
5. Interactive visualization and interrogation foundations of EM
This section will present foundations for visualization and interrogation
of Extended Matrices at runtime. One of the main goals within
visual data mining is to present the whole ontological richness of the
EM (relationships between SUs, chronological horizons of reconstruction
hypothesis, sources behind scientific reasoning, etc.). First we will
introduce a minimal set of requisites to efficiently represent and visualize
an Extended Matrix, data granularity aspects and common
bottlenecks typically encountered during the engineering of specific
interrogation tools. We will stress the advantages of 3D visual inspection
of Extended Matrices applied to interactive and collaborative workflow.
We will then provide a set of best practices, developed for the
design and implementation of consistent algorithms for real-time extraction
and visualization within virtual environments. Such guidelines
abstract from the actual 3D framework or software library adopted,
except for the assumption that basic scene-graph functionalities have to
be provided by the chosen framework.
5.1. Introduction and requisites
One of the main requirements for functional inspection and interrogation
within the Extended Matrix Framework is a clear separation
between the visible scene-graph and query-able graph (or ProxyGraph),
as proposed in previous work and researches (Tobler, 2011,
Serna et al., 2012). Such approach offers great advantages and flexibility
when dealing with different data granularity between the two
graphs: specifically it separates the 3D graphics requirements for visible
scene-graphs (multi-resolution, hierarchical culling, cascading transformations,
etc.) from semantic segmentation requirements. This avoids
conceptual pollution between two really different representations and
domains. For instance, even if a 3D floor element can be modelled by a
simple geometry (i.e: a single node) within the visible scene-graph, it
may present a richer and finer Proxy-Graph for semantic interrogation
(a 3D floor can have different phases and restorations). On the other
hand, a complex scene-graph employed to represent a 3D element - for
instance, a very detailed, multi-resolution column through hierarchical
level-of-details - may map to a single proxy-node. Such requirement is
even enforced if we deal with procedural scene-graphs, for instance, a
columnade generated by instancing rules where just one column is
query-able. Complexity is even increased when we deal with the
mapping of multi-temporal virtual environments: an Extended Matrix
spans across multiple time periods, thus leading to multiple scenegraphs,
each per period. Consequently, another requirement is indeed
to provide mechanics for visualizing and switching multiple scenegraphs
representing the same context at different temporal slices. Once
again, the hierarchy used for a 3D acquisition of a specific
Source-based model
1150A.D.
Source-based model
1100A.D.
Reality-based model
2014A.D.
a)
b)
c)
Fig. 12. From a reality-based model it is possible to make several reconstruction hypotheses:
ideally at least one for each period formalized in the Matrix of Harris.
D.01
Fig. 13. Time management in the EM (1.1). Multiple phases are eventually integrated
with 3D models snapshots as in the case of the 2nd A.D. horizon. The image represents an
extract from the EM of the Great Temple (see Fig. 11). At present a reconstruction has
been proposed only for the 2nd A.D. epoch.
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
508
archaeological site may dramatically differ from the one used for the 3D
reconstruction of the same site. To obtain a smooth and efficient pipeline,
combined with interactive and collaborative aspects dealing
with inspection, visualization and query of a single Extended Matrix,
the following are generally required:
• An Extended Matrix (EM), described through means of GraphML
format
• Resource folders containing 3D proxy elements, expressed by
common 3D formats
• Multiple visible 3D scene-graphs, for each time-period expressed in
EM
• A cloud storage (optional) for all the above data, providing a distributed
workflow
The following sections will describe the blueprint and necessary
concepts in order to craft correct and interactive 3D inspection tools
dealing with Extended Matrices.
5.2. The EMviq tool
In order to validate described foundations in this section and provide
a complete visualization and interrogation context within the
framework, a 3D inspection tool “Emviq” was developed. The implementation
focuses on extraction correctness (described in the next
section), ease-of-use and performance, in order to establish a fast and
robust pipeline within the EM. The interactive tool allows users to define
a collection of Extended Matrices (by providing paths or URLs to
GraphML files, geolocation and other attributes) and a collection of
scene-graphs covering different time periods described by the main EMtimeline
(see Section 5.5). Since the collaborative and iterative nature
of EM pipeline, a cloud-based approach is employed to allow smooth EM
design and consequent visual interrogation sessions, in order to validate
or debug current Extended Matrix. Furthermore, the cloud enables
different professionals to design and operate remotely on multiple Extended
Matrices and Scene-graphs for each time-period, including reconstruction
hypotheses.
5.3. Extraction and 3D translation
A single Extended Matrix (EM) can be described by a GraphML file,
readily editable through existing visual tools offering flexible and
iterative workflow between design and visualization phases. To inspect
and query such data within a 3D interactive context, we need to perform
a translation of the EM into appropriate, queryable structures,
suitable for 3D interactive interrogation. First of all, the following need
to be extracted or computed - more precisely - every time an Extended
Matrix is modified or updated:
1. Source-Graphs: an internal runtime representation of EM sources
relationships (paradata)
2. The Proxy-Graph: a graph for real-time queries and interrogation,
handling 3D proxies objects well defined in 3D space
3. A Timeline (or EM-Timeline): a finite number of time-periods,
identified by beginning and end values
Source-graph, Proxy-graph and Timeline can be extracted using any
GraphML or XML parser, by taking into account the basic graph-topology
(nodes and edges) defined for the GraphML format (Brandes
et al., 2010). Such extraction step needs to be performed only when the
involved EM is modified, more precisely, it has to be performed only on
the modified EM sub-graphs. Source-graphs are the induced sub-graphs
of the Extended Matrix obtained by filtering dotted edges (paradata).
The Proxy-Graph instead, is generated from the EM and a given set of
3D shapes (3D Proxies) as proposed in Demetrescu (2015) and Serna
et al. (2012), with each proxy-node identified by a unique ID. The EMnodes
that will be realized as proxies are: Special Find nodes, SU nodes,
structured and non-structured virtual SU nodes and Seriation nodes. The
Seriation node in particular, generates a sub-graph by procedurally
instancing the referenced 3D proxy element into multiple locations,
depending on transformation rules: these are provided as list of transformation
matrices (translation, rotation and scale) for that specific 3D
proxy element. A typical application for a Seriation node are for instance
columnades, generated from a single proxy column (see Proxygraph
in Fig. 14).
For a given Extended Matrix and a starting EM-Node, Algorithm 1
recursively build a Proxy-graph G. A RealizeProxyGeometry() procedure
will first realize the proxy geometry for current input EM-node e: this
will also take care of the special Seriation-node, by procedurally instancing
the proxy-geometry if a correspondent transformation-list is
found. Finally it will recursively proceed with its children and return
the finalized Proxy-graph.
Fig. 14. An Extended Matrix (left) from which the following are automatically extracted (right, from top to bottom): Proxy-graph (A) including realized seriation-nodes for procedural
column proxies, Source-graphs (B) and Timeline (C).
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
509
Algorithm 1. Realize and return Proxy-Graph, starting from an
Extended Matrix node e
The realized Proxy-graph is actually the query-able 3D skeleton of
an Extended Matrix, spatially plunged in a virtual environment. When
coupled with a visible Scene-graph, the Proxy-graph is typically hidden
and only perceivable by 3D queries, except for specific highlighting
requirements. Each node in the Proxy-graph has the following attri-
butes:
• A unique identifier, mapping a specific EM-node (e.g.: “SU010”)
• The node type: Special Find, SU, structured and non-structured
virtual SU or Seriation
• URL, Description and Label
• A temporal value: in the Extended Matrix this corresponds to vertical
axis (y) - see Fig. 14. Through the extracted EM-Timeline, this
value also maps to a specific time-span or Time period (e.g.:
“2ndAD”)
• The EM ID the proxy-node belongs to
On the other hand, each node in each Source-Graphs (paradata) has
the following attributes:
• A unique identifier, mapping a specific EM-node (e.g.: “SU010”)
• A set of direct Source-graph children
After the realization of Proxy-Graph and Source-Graphs, the interactive
tool can access a proxy-node in the 3D virtual space through a 3D
query: through the unique ID we have direct access to its Source-graph,
its type, its time period, labelling information and of course which
Extended Matrix we are interrogating.
5.4. Interactive 3D interrogation
An interactive interrogation is spatially performed on a Proxy-Graph
or even several Proxy-Graphs, when we deal with multiple Extended
Matrices. This task is carried out by the interactive tool by fast intersection
routines (the 3D querier) - typically offered by all modern 3D
frameworks and libraries - to retrieve all needed information previously
described. Since the simplified nature of proxy-geometries in fact, such
routines are performed quickly against basic 3D shapes. More importantly,
they totally abstract from the complexity of the visible Scenegraph,
potentially presenting very complex geometries. The 3D query is
performed from a specific position in virtual space (typically the user
camera, but not limited to) with a given direction (typically camera
target or a custom one).
An issue that may occur when rich or complex Proxy-graphs are
defined, is query occlusion. This basically causes some proxy geometry
to occlude each other (for instance, when multiple SU-proxies are
stacked) thus making some queries partially or fully inaccessible at
runtime. This can be solved by providing query peeling: this mechanism
makes it possible to temporarily “peel off” or carve proxies with a given
policy, to reach otherwise inaccessible ones. The policy is typically
based on camera position and an interactive radius (spherical peeling see
Fig. 15), although other approaches can be employed, depending on
specific user or application needs (e.g.: camera-independent peeling).
Fig. 15. An example of interactive spherical 3D peeling that makes it possible to perform 3D queries on occluded or partially occluded proxies. User-controlled radius provides visual
interrogation on nested proxies.
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
510
The query peeling is often coupled with a corresponding 3D cut applied
to the visible Scene-graph, to offer a clear visual feedback.
5.5. Handling temporal dimension
We already introduced the Timeline, extracted from an EM (see
Section 5.3) and requirements to handle multi-dimensional virtual environments:
the latter is realized by means of multiple Scene-graphs,
each covering a time-period (e.g.: “2ndAD” will map a specific visible
Scene-graph) all representing the same context. This section will define
management of temporal warping and visualization of multiple timeperiods.
For a given Extended Matrix and its Timeline, we define a
direct mapping from a generic EM-node e to a time-period τ:
→T e τ( )
where τ is a unique identifier (e.g.: “XX century”) that allows us to
identify a Scene-graph for a given EM-node. A selector S can be deployed
at runtime to activate on-demand temporal-warp (switch) of the
visible Scene-graph s depending on input period τ:
→S τ s( )
Within an interactive inspection, such selector can be easily attached
to a proxy-node and triggered by specific event (e.g.: some user
input), thus elevating such nodes to temporal hyperlinks in a 3D virtual
environment (see Fig. 16). Generally, the number of Scene-graphs
matches the number of time-periods, although different time-periods (τ,
τ′, τ″, ...) may refer to the same Scene-graph or a part of it. Especially in
the latter case, the hierarchical approach offers huge flexibility in terms
of scene design. It is quite common in fact, that a single Scene-graph s
may contain a sub-graph that is shared with another time-period (thus a
scene portion re-used by another graph s′). This kind of cross-temporal
organization, allows elegant and compact scene design (see Fig. 17).
It is important to highlight the abstraction from number or complexity
of visible scene-graphs (s,s′,...) and duration of time periods: a
given τ can in fact represent an entire era, a single year, one day or even
a few minutes. Such level of abstraction well suits Extended Matrix
requirements and multi-temporal visualization needs. During interactive
fruition in fact, proxy-graph itself is exploited as temporal
medium and interface to explore current context at different periods.
5.6. Presentation of Source-Graphs
During interactive interrogation of a Proxy-graph, it is often required
to visually present data relationship associated with a specific
proxy-node. As previously introduced, such information can be retrieved
from Source-graphs, although different solutions can be employed
to present and explore such semantic data through clear layouts
in a 3D space. Previous research has already dealt with appealing representation
and visualization of hierarchies and linked data, especially
in 2D contexts (Book and Keshary, 2001 Pavlo et al., 2006 Mazumdar
et al., 2015). Within interactive visualization and query of extended
matrices, parent-centered layouts are typically the most appropriate
solution, although the third dimension play an important role into the
layout algorithm, especially when the interactive query is performed
through immersive visualization systems (Head-mounted Displays,
CAVE systems, etc.).
Within extended matrices, the main guideline for visual presentation
of Source-graphs is to create a set of 3D radial-graphs - for each
proxy-node - encapsulating proxy-centered source-graphs and connected
information, shown upon user interrogation. The main goal is to produce
a consistent but non-intrusive 3D layout (see Fig. 18), exploiting
and taking advantage of third dimension within standard or immersive
VR sessions, thus aiding and enhancing interrogation of complex da-
tasets.
6. Change-log for the EM 1.1 - EMF 1.1
The version 1.1 of the Extended Matrix adds a few tools and
Fig. 16. A proxy-node (USV132) exploited as temporal hyperlink from current period (today) back to its connected time period (2nd A.D.) and its reconstruction hypothesis.
Fig. 17. A sample visible scene-graph s (right) shares sub-graphs A and B with another
scene-graph s′ (left): A and B consequently span across two different time periods τ and τ′
(respectively S(τ) → s and S(τ′) → s′) through cross-temporal instancing.
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
511
conventions for the integration with a 3D representation.
EM 1.1 version:
• Proxies and representation models: they allow coherent visualizations
of a reconstruction hypothesis in the 3D space (for details see examples
in Section 4.3).
• Snapshots: a snapshot is a survey of a model collected in a precise
moment in time (i.e. the photogrammetric survey of the Great
Temple in 2014). The snapshots are palimpsest of a context. In order
to isolate the elements pertaining to a given epoch they have to be
geometrically segmented and cleaned by posthumous portions.
• Graphical and textual documentation for scientific publication of a reconstruction
hypothesis (EM): EM with subdivision in virtual activities,
Report of Virtual Activities, Dossier comparatif, lists of EM
nodes grouped according to the virtual activities, sections and plans
of the proxy model (per-epoch).
• New conventions in node connection (archs) to simplify and made
clearer the visual representation of the EM have specific documentation
(whitepapers, on-line wiki, etc.).
EMF 1.1 version:
• Emviq: it connects the 3D models (reconstructive and snapshots)
with the reconstruction hypothesis (EM). It can be used as a convenient
test for the formal correctness of the EM (see Section 5).
7. Conclusions
This contribution adds full support to 3D representation (see
Sections 4 and 4.3) of the Extended Matrix (ver. 1.1). A real case scenario
(the Great Temple of the ancient Roman town Colonia Dacica
Sarmizegetusa) is used in order to make clear the introduced notions
and validate the adopted formal solutions.
Within the Extended Matrix Framework (EMF), an interactive tool
(“Emviq”) was designed and developed to validate visual foundations
described in Section 5. Resulting blueprints include: algorithms and
formalisms for automatic generation of 3D Proxy-graphs starting from
EM descriptions, design of cross-temporal scene-graph, techniques for
expert inspection, interactive 3D query and scalable visualization of
graph data in multi-temporal virtual contexts. Important outcomes of
the research include:
• An iterative workflow within the EM that involves data collection,
visual validation and debug of EM description, simplification of data
sharing between researchers with different background;
• Blueprints for robust 3D interactive inspection and query of Extended
Matrices, applied to the development of visual tools that focus on 3D
presentation (including immersive fruition) of complex information
and adoption of cloud-based approaches for automatic data aggregation
and runtime 3D validation;
• A data driven approach from reconstructive record (EM) to 3D
Fig. 18. An example of parent-centered 3D layout to present
Source-graphs upon interactive interrogation. Each
node 3D position is computed depending on graph hierarchy,
user position, distance, node content and spatial
extents.
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
512
depiction that encourages quick information exchange and interactive
workflow on dynamic 3D scenes;
While EM stimulated the development of different tools and technical
solutions, on the other hand, it does not focus on the technologies
employed. It offers specialized semantics for description of multidimensional
elements within a visual approach: (A) a set of nodes and
relationships to describe the reconstruction process using a schema-less
approach and (B) blueprints for interactive 3D query and exploration of
multidimensional graph databases including spatial inspection, complete
data retrieval and coherent navigation of temporal dimension. EM
does not focus on the description of the digital asset but aims to a detailed
documentation of a “potential context” perceived as if it were still
existing in front of the specialist in charge of the reconstruction. In this
way the digital instance is just one of the properties of the reconstructed
context along with the other properties declared in the EM database.
7.1. Future works
The development of the EM and its related EMF will result in new
versions (1.2, 1.3, etc.) with the addition of both methodological and
technical improvements. The next steps will be:
• EM. Support for different, self excluding reconstruction hypotheses
(as a 5th dimension for the EM). In some cases there are more than
one possible reconstructive solution that have to be stored and organized
accordingly in the EM.
• EM. A classification of standard scenarios where EM can be employed
is work in progress. It could clarify the limits of applicability
in certain typologies of context.
• EMF. Use of server side graph database solutions (like, as an example,
Neo4j): they are a fast-growing trend that enables smart data
aggregation and mining on complex and large scale scenarios.
• EMF. Another future step, already in progress, involve porting of
Emviq to WebGL, on top of multiple scene-graph based javascript
libraries. Such process will enable on-line 3D fruition of Extended
Matrices on all major browsers without installing any additional
plug-in or software.
References
Apollonio, F.I.F.I., 2016. Classification schemes for visualization of uncertainty in digital
hypothetical reconstruction. In: 3D Research Challenges in Cultural Heritage II.
Springer, pp. 173–197.
Apollonio, F.I.F.I., Gaiani, M.M., Benedetti, B.B., 2012, May, May. 3d reality-based artefact
models for the management of archaeological sites using 3d Gis: a framework
starting from the case study of the Pompeii Archaeological area. J. Archaeol. Sci. 39
(5), 1271–1287.
Apollonio, F.I.F.I., Giovannini, E.C.E.C., 2015. A paradata documentation methodology
for the Uncertainty Visualization in digital reconstruction of CH artifacts. SCIRES-ITSCIentific
Res. Inform. Technol. 5 (1), 1–24.
Ascani, K., Kermit, H., Skytte, G., 2002. In: Stenone, Niccolo (Ed.), Anatomista, geologo,
vescovo. Atti del seminario organizzato da Universitetsbiblioteket i Troms e
l'Accademia di Danimarca (Roma 23 ottobre 2000). Analecta Romana Instituti Danici
Suppl. 31. Roma.
Auer, M.M., Agugiaro, G.G., Billen, N.N., Loos, L.L., Zipf, A.A., 2014. Web-based visualization
and query of semantically segmented multiresolution 3d models in the field
of cultural heritage. ISPRS Ann. Arch. Photogramm. Remote. Sens. Spat. Inf. Sci. 2
(5), 33.
Baldwin, T.D.T.D., Flaten, A.R.A.R., 2011. Adapting the Agile Process to Digital
Reconstructions of the Temple of Apollo at Delphi. Revive the Past, 30.
Book, G.G., Keshary, N.N., 2001. Radial tree graph drawing algorithm for representing
large hierarchies. University of Connecticut.
Brandes, U.U., Eiglsperger, M.M., Lerner, J.J., Pich, C.C., 2010. Graph markup language
(GraphML). Citeseer.
Bruschke, J.J., Wacker, M.M., 2016. Simplifying documentation of digital reconstruction
processes. In: Münster, S.S., Pfarr-Harfst, M.M., Kuroczyński, P.P., Ioannides, M.M.
(Eds.), 3D Research Challenges in Cultural Heritage II: How to Manage Data and
Knowledge Related to Interpretative Digital 3D Reconstructions of Cultural Heritage.
Springer International Publishing, Cham, pp. 256–271. http://dx.doi.org/10.1007/
978-3-319-47647-6_12.
Bruseker, G.G., Guillem, A.A., Carboni, N.N., 2015. Semantically documenting virtual
reconstruction: building a path to knowledge provenance. ISPRS Ann. Photogramm.
Remote. Sens. Spat. Inf. Sci. 1, 33–40.
Dell’Unto, N.N., Leander, A.A., Dellepiane, M.M., Callieri, M.M., Ferdani, D.D., Lindgren,
S.S., 2013. Digital reconstruction and visualization in archaeology: case-study drawn
from the work of the Swedish Pompeii Project. In: Proceedings of the DigitalHeritage
2013 - Federating the 19th Int’l VSMM, 10th Eurographics GCH, and 2nd UNESCO
Memory of the World Conferences, Plus Special Sessions fromCAA, Arqueologica 2.0
et al. Vol. 1. IEEE, Marseille, France, pp. 621–628. http://www.scopus.com/inward/
record.url?eid=2-s2.0-84896763525&partnerID=40&md5=
a14816775bd754d0c3c6fd760b6f1450.
Demetrescu, E.E., 2015. Archaeological Stratigraphy as a formal language for virtual
reconstruction. Theory and practice. J. Archaeol. Sci. 57, 42–55. http://www.
sciencedirect.com/science/article/pii/S0305440315000382.
Demetrescu, E.E., Ferdani, D.D., Dell’Unto, N.N., Touati, A.-M.L.A.-M.L., Lindgren, S.S.,
2016. Reconstructing the original splendour of the House of Caecilius Iucundus. A
complete methodology for virtual archaeology aimed at digital exhibition. SCIRES-ITSCIentific
Res. Inf. Technol. 6 (1), 51–66.
Doerr, M.M., 2003, Sep, Sep. The CIDOC conceptual reference module: an ontological
approach to semantic interoperability of metadata. AI Mag. 24 (3), 75.
Doerr, M.M., Kritsotaki, A.A., Rousakis, Y.Y., Hiebel, G.G., Theodoridou, M.M., 2014.
CRMsci: The Scientific Observation Model.
Doerr, M.M., Theodoridou, M.M., 2011. CRMdig: A Generic Digital Provenance Model for
Scientific Observation. Proc. TaPP 2011, 3rd, 00006.
Forte, M.M., Dell’Unto, N.N., Issavi, J.J., Onsurez, L.L., Lercari, N.N., 2012. 3D
Archaeology at Catalhoiük. Int. J. Herit. Digit. Era 1 (3), 351–378.
Gonzalez-Perez, C.C., Martín-Rodilla, P.P., Parcero-Oubiña, C.C., Fábrega-Álvarez, P.P.,
Güimil-Fariña, A.A., 2012. Extending an abstract reference model for transdisciplinary
work in cultural heritage. In: Metadata and Semantics Research. Springer, pp.
190–201.
Gros, P.P., 1995. Hercule à Glanum: Sanctuaires de transhumance et développement
urbain. Galliapp. 311–331.
Hauck, O.O., Kuroczyński, P.P., 2014. Cultural Heritage Markup Language.
Kuroczyński, P.P., Hauck, O.O., Dworak, D.D., 2014. Digital Reconstruction of Cultural
Heritage–Questions of Documentation and Visualisation Standards for 3D Content.
In: 5th International Euro-Mediterranean Conference on Cultural Heritage, Lemessos/
Cyprus.
Kuroczyński, P.P., Hauck, O.O., Dworak, D.D., 2016. 3D Models on triple paths - new
pathways for documenting and visualizing virtual reconstructions. In: Münster, S.S.,
Pfarr-Harfst, M.M., Kuroczyński, P.P., Ioannides, M.M. (Eds.), 3D Research
Challenges in Cultural Heritage II: How to Manage Data and Knowledge Related to
Interpretative Digital 3D Reconstructions of Cultural Heritage. Springer International
Publishing, Cham, pp. 149–172. http://dx.doi.org/10.1007/978-3-319-47647-6_8.
Landeschi, G.G., Dell’Unto, N.N., Lundqvist, K.K., Ferdani, D.D., Campanaro, D.M.D.M.,
Touati, A.-M.L.A.-M.L., 2016. 3D-GIS as a platform for visual analysis: investigating a
Pompeian house. J. Archaeol. Sci. 65, 103–113.
Mazumdar, S.S., Petrelli, D.D., Elbedweihy, K.K., Lanfranchi, V.V., Ciravegna, F.F., 2015.
Affective graphs: the visual appeal of linked data. Semantic Web 6 (3), 277–312.
Medri, M.M., 2003. Manuale di rilievo archeologico. Roma-Bari.
Münster, S.S., 2013. Workflows and the role of images for virtual 3d reconstruction of no
longer extant historic objects. ISPRS Ann. Photogramm. Remote. Sens. Spat. Inf. Sci.
197–202 II-5/W1.
Münster, S.S., Hegel, W.W., Kröber, C.C., 2016a. A model classification for digital 3d
reconstruction in the context of humanities research. In: Münster, S.S., Pfarr-Harfst,
M.M., Kuroczyński, P.P., Ioannides, M.M. (Eds.), 3D Research Challenges in Cultural
Heritage II: How to Manage Data and Knowledge Related to Interpretative Digital 3D
Reconstructions of Cultural Heritage. Springer International Publishing, Cham, pp.
3–31.
Münster, S.S., Kröber, C.C., Hegel, W.W., Pfarr-Harfst, M.M., Prechtel, N.N., Uhlemann,
R.R., Henze, F.F., 2016b. First Experiences of Applying a Model Classification for
Digital 3d Reconstruction in the Context of Humanities Research. In: Ioannides, M.M.,
Fink, E.E., Moropoulou, A.A., Hagedorn-Saupe, M.M., Fresa, A.A., Liestøl, G.G.,
Rajcic, V.V., Grussenmeyer, P.P. (Eds.), Digital Heritage. Progress in Cultural
Heritage: Documentation, Preservation, and Protection: 6th International
Conference, EuroMed 2016, Nicosia, Cyprus, October 31– November 5, 2016,
Proceedings, Part I. Springer International Publishing, Cham, pp. 477–490. http://dx.
doi.org/10.1007/978-3-319-48496-9_37.
Münster, S.S., Niebling, F.F., 2016. Building a Wiki Resource on Digital 3D Reconstruction
Related Knowledge Assets.
Opitz, R.S.R.S., Johnson, T.D.T.D., 2016. Interpretation at the Controller's Edge: designing
graphical user interfaces for the digital publication of the excavations at Gabii
(Italy). Open Archaeol. 1 (1).
Pavlo, A.A., Homan, C.C., Schull, J.J., 2006. A Parent-centered Radial Layout Algorithm
for Interactive Graph Visualization and Animation. arXiv preprint cs/0606007.
Pfarr-Harfst, M.M., 2016. Typical workflows, documentation approaches and principles of
3D digital reconstruction of cultural heritage. In: Münster, S.S., Pfarr-Harfst, M.M.,
Kuroczyńki, P.P., Ioannides, M.M. (Eds.), 3D Research Challenges in Cultural
Heritage II: How to Manage Data and Knowledge Related to Interpretative Digital 3D
Reconstructions of Cultural Heritage. Springer International Publishing, Cham, pp.
32–46. http://dx.doi.org/10.1007/978-3-319-47647-6_2.
Pfarr-Harfst, M.M., Grellert, M.M., 2016. The reconstruction - argumentation method. In:
Ioannides, M.M., Fink, E.E., Moropoulou, A.A., Hagedorn-Saupe, M.M., Fresa, A.A.,
Liestøl, G.G., Rajcic, V.V., Grussenmeyer, P.P. (Eds.), Digital Heritage. Progress in
Cultural Heritage: Documentation, Preservation, and Protection: 6th International
Conference, EuroMed 2016, Nicosia, Cyprus, October 31–November 5, 2016,
Proceedings, Part I. Springer International Publishing, Cham, pp. 39–49. http://dx.
doi.org/10.1007/978-3-319-48496-9_4.
Potenziani, M.M., Callieri, M.M., Dellepiane, M.M., Corsini, M.M., Ponchio, F.F.,
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
513
Scopigno, R.R., 2015. 3dhop: 3D heritage online presenter. Comput. Graph. 52,
129–141. http://www.sciencedirect.com/science/article/pii/S0097849315001041.
Ronzino, P.P., Niccolucci, F.F., Felicetti, A.A., Doerr, M.M., 2015. CRMba a CRM extension
for the documentation of standing buildings. Int. J. Digit. Libr. 1–8.
Rumbaugh, J.J., Jacobson, I.I., Booch, G.G., 2004. Unified Modeling Language Reference
Manual, The. Pearson Higher Education.
Serna, S.P.S.P., Schmedt, H.H., Ritz, M.M., Stork, A.A., 2012. Interactive semantic enrichment
of 3D cultural heritage collections. In: VAST: International Symposium on
Virtual Reality, Archaeology and Intelligent Cultural Heritage, pp. 33–40.
Shi, W. e. A.W. e. A., 2016. CHER-Ob: a tool for shared analysis in cultural heritage. In: .
Tobler, R.F.R.F., 2011. Separating semantics from rendering: a scene graph based architecture
for graphics applications. Vis. Comput. 27 (6-8), 687–695 00007.
von Schwerin, J.J., Lyons, M.M., Loos, L.L., Billen, N.N., Auer, M.M., Zipf, A.A., 2016.
Show me the data!: structuring archaeological data to deliver interactive, transparent
3D reconstructions in a 3D WebGIS. In: Münster, S.S., Pfarr-Harfst, M.M.,
Kuroczyński, P.P., Ioannides, M.M. (Eds.), 3D Research Challenges in Cultural
Heritage II: How to Manage Data and Knowledge Related to Interpretative Digital 3D
Reconstructions of Cultural Heritage. Springer International Publishing, Cham, pp.
198–230. http://dx.doi.org/10.1007/978-3-319-47647-6_10.
E. Demetrescu, B. Fanini Journal of Archaeological Science: Reports 14 (2017) 500–514
514