An Information System to Analize Cultural Heritage Information

An Information System to Analize Cultural Heritage Information

An Information System to Analize Cultural
Heritage Information
J.C. Torres1 , L. López1 , C. Romo1 , and F. Soler1
Virtual Reality Laboratory. University of Granada, Spain
[email protected],
WWW home page: http://lrv.ugr.es/chis
Abstract. Managing information related to cultural heritage sites is an
important task and much work has been devoted to developing special
purpose document management systems. These systems are able to store
and retrieve large amounts of documents; however, while this is adequate
for some purposes, it is not sufficient for research and conservation work.
Researchers need to determine relationships between data, and the most
important relationships in cultural heritage information are spatial relationships.
A new kind of information system is therefore needed, in which the 3D
representation of an object is a blackboard on which all data is represented. This paper proposes the concept of Cultural Heritage Information
Systems, and presents our implementation of the system. An example
application illustrating the use of the system is also presented.
1
Introduction
Archaeologists and other professionals in the field of Cultural Heritage (scientists, curators, restorers, architects, and so on) manage a large amount of information related to heritage objects. This information includes a wide variety of
elements, for example: photographs, paintings, maps, descriptive texts, historical
documents, annotations, measurements or test results.
In our opinion, although traditional information systems are very useful for
managing large amounts of data, they have two important drawbacks when used
for Cultural Heritage applications:
– For most information, position is a relevant attribute. Although position
attributes can be attached to any database record indicating from which
part of the monument the piece of data was collected, this will not work
when several names are used for the same place. Moreover, the data may be
related to any area, such as the part of a wall that has been treated during
a restoration process.
– Document management systems do not allow computations related to the
location of information to be performed. All queries performed in the system
are textual, and consequently the user obtains a list of relevant documents.
However, scientists and other cultural heritage professionalsmay be also interested on queries involving the spatial distribution of the information.
Fig. 1. Screenshot from the application showing the head of a lion sculpture from the
Court of Lions at the Alhambra Palace.
In order to make up for these deficiencies, the system should be able to
relate the managed information with its spatial location as well as implement a
mechanism to perform queries based on the information location. Additionally,
the query results should be shown as spatial information.
Geographical Information Systems essentially work following this approach.
However GIS can not be used for movable heritage items nor element that are
not deployed horizontally (as walls or façades).
In this paper we present a new software tool (named a Cultural Heritage
Information System) which allows information layers to be attached to the surface of any cultural heritage artefact. Layers can be rendered, edited, copied,
queried interactively or combined, in the same way as a GIS. Figure 1 shows a
screenshot of our implementation of the system (Chisel). Our system is not a
GIS, and does not use a GIS, it performs like a GIS.
The rest of this paper is structured as follows: Section 2 presents previous
work, Section 3 introduces the basic ideas upon which our system has been
constructed and explains its basic functionality, and Section 4 describes a simple
example of its application.
2
Previous work
GIS has become an standard for archaeological research. A comprehensible review of its use and a case study can be found in [Kata09]. But, as have been
argued on the previous section they are not adequate to study artefacts.
Data management systems have been successfully used to organize cultural
heritage repositories, on which the main goal is to be able to retrieve an artefact
from some of their metadata. One remarkable work on this area has been done
by Felicetti and Niccolucci [Feli11], who presented a management system build
using open source components. This approach can be used basically to browse
and search element collections.
Many attempt have been made to add annotations and external information
to 3D model. Most of them require to segment the model, that is to decide to
which section of the model it is possible to link the information. Very frequently
this process must be done as a previous process, preventing to create new association dynamically. This approach is appropriate for museum, or for systems
whose purpose is to show a previously processed information, but not for research. For instance, ViSMan is an open-source visualization framework that
has been used for virtual reconstructions and data management in archaeology.
Its allows to link external documents to 3D landscapes enabling its conceptualization [Diam10].
Perhaps the most flexible proposal of a segmented labeling system has been
done by an international group including VCL from Italy. Their system allows
to do annotation on different multimedia objects, using an uniform way to define
areas on different multimedia objects [Pena11].
3
Structure of the Cultural Heritage Information System
The main difficulty when designing a system that associates information layers
with the surface of an object is to establish mapping applications between layers
representation and surface points. While GISs use projections to perform this
mapping, this is not possible for our system as the geometry of every artefact will
be different and their boundaries may constitute complex surfaces. The mapping
for Chisel is obtained by partitioning the surface of the artefact into cells and
assigning an unique integer identifier to each one. This allows information layers
to be represented as sequence of attribute values. Each cell is assigned the value
that is stored at its corresponding position in the sequence.
Figure 2 shows a 2D diagram of the partitioning process. The space intersected by the surface is divided into a set of non overlapping cells of equal size.
The interior part of the object is not indexed. The cells are assigned a unique
identifier (figures inside the cells) and layers are stored as a sequence of values.
Cell identifiers can be assigned in any order, but it is important that they are
fixed for each object. This allows data to be assigned to the surface and make
correspondences between different information layers. The diagram in figure 2 is
2D. In a real 3D case the cells are 3D boxes, but anyway identifiers are integers.
This means that layers are a one dimensional arrays of values.
It should be noted that although the cells are all the same size, the area of the
surface within each cell is different. That is, cell size determines the maximum
area of the surface that will be assigned a value.
This representation allows raster layers to be managed flexibly, in a similar
way to a GIS raster map. The simplicity of the representation allows to implement complex operations in a simple way.
Fig. 2. Assigning identifiers to cells allows property layers to be defined as a onedimensional array of values.
Layers can hold null values, implemented as a bits array, with the same size
of the layer array, indicating whether the stored value is valid. It is possible to
associate layers with a database table in order to manage non-numeric data. In
this situation, the layer array contains the primary key for the record associated
with the cell. The application can deal with text, dates, numbers, images, films
and any other document as a database record field.
4
System functionality
This section describes the functionality of the system, providing a general overview
of its capabilities.
Creation of the models The system has been designed to create representations of artefacts from digital models generated using a laser scanner. It can read
a 3D model stored using the PLY format, an open format that was specifically
designed to store three dimensional data of 3D scanned objects.
The application can automatically import the surface description of an artefact from a PLY file, computing its cell decomposition and its topological relationship. The user must specify the size for the cells. The prototype can manage
representations of objects with a cell size of 2 mm for an object whose dimensions
are over 1 km.
The representation of the Chisel models contains three connected components: geometry, data layers and database. When the model is created from
an external file it contains only geometric information. The user can create information layers in several ways: interactively, from geometric information or
from previously created layers. Figure 3 shows an overview of the application
structure.
Fig. 3. Schematic overview of the system. The representation of the Chisel model
contains three connected components: geometry, data layer and database.
Layer rendering Layers are rendered as colour applied to the geometric model.
The user assigns a colour table to every layer, and is able to select the ordered
sequence of layers that will be rendered at any moment as well as whether the
original model texture will be visible under the layer colour.
Queries Users can obtain the property value assigned to any surface point,
on any given layer, by simply clicking on the point. The input information is
a surface point, and the output information is the associated record on the
database. This kind of query operation is similar to that which is available on
most labelled 3D models.
Our system also allows SQL queries to be made, whose results are shown as
3D visualizations of the set of points satisfying the query condition. In this case,
the input is a condition and the output is a subset of the surface.
To illustrate this, Figure 4 shows the result of a query on the ceiling of the
“Hall of the Kings” in the Alhambra Palace, where a search was made for areas
restored by ‘Ramón’.
Layer operations. Attribute layers can be combined and transformed generating new layers. These operations are used to analyze the information. Chisel
includes the following operations:
Transformations: operations that apply a function to the attribute values (in
other words performing a recodification of the attributes).
Mathematical functions: the powerful r.mapcalc operation defined in GRASS
GIS has been implemented in Chisel [Nete08]. This allows new layers to be
generated by specifying cell values using arithmetic and logical operations on
previously defined layers.
Null manipulation: null cells can be assigned a value, and null values can be
assigned to cells with specified values.
Fig. 4. Interactive query on a model of the ceiling of the“Hall of the Kings” in the
Alhambra Palace.
Interactive edition. The most obvious way to introduce information to the
system is to create information layers. Layers can be created interactively describing the structure of the associated database table and editing the layer values on the 3D model. The editing process comprises an interactive “paint-like”
operation: in other words, the user selects the areas of artefact with a specified
value by simply “painting” them with a brush.
Geometric layers. Analyses can also be made involving geometry. To do this,
Chisel includes operations to compute information layers from geometry:
Curvature. This function generates a layer by assigning a curvature value to
each cell. Curvature is computed as the inverse of the average radius of tangent
spheres on the surface at the cell and the surrounding cells.
Roughness. The resulting layer contains the average deviation of neighboring
cells to the tangent sphere for each cell.
Distance field. The layer assigns to each cell its distance (along the surface)
to the non null cells of a given layer or to a point on the surface.
Orientation. The output layer represents the orientation angle between the
normal vector at the cell and a given reference vector.
Normal. The output is a set of layers containing the components of the normal
vector at every cell.
Reports. Most of the operations that can be performed generate a new layer
containing a result which can be rendered on the 3D model. This is convenient
for showing distribution values, but sometimes quantitative results are needed.
Report tool generates information about the surface that is covered by each
attribute value.
Although this operation accepts only one layer as input, complex analyses can
be performed which combine more than one layer using the previously described
operations.
5
Case study
We present here a simple application example working with a fragment of an
Iberian vessel (see Figure 5). This piece has been restored from several fragments that was glued together. The example try to analyse the curvature of the
fragments border in order to study its erosion. The 3D model was acquired by
scanning the object using a Konica Minolta Vivid-910 laser scanner. Using this
device we were able to obtain a triangular mesh with an accuracy in the order of
tenths of millimeters. Using this geometric information as input, the application
generated a 3D object with the cells size of 0.05 mm (See figure 5 top left).
Fig. 5. Case study: Analysis of the edge curvature of an Iberian vessel. See section 5
for explanation.
Once the piece has been loaded by Chisel we create a segmentation layer that
identifies every fragment. That is the category for any cell is the identifier of its
fragment (Figure 5 top center).
A second layer is created interactively identifying cell that are on the fragments edge (Figure 5 top right). From this layer a buffer is created extending
the edges to the adjacent areas. This has been done computing a distance field
from the edges and selecting the cells whose distance is bellow 0,5 mm.
The buffer layer is combined with the previously created fragments layer, creating a fragment edge layer. Its cells have the fragment identifier if they are on the
edges, and null otherwise (Figure 5 bottom left). This combination is done using
an algebraic expression: EdgesID = if (isnull(buf f er), null(), f ragment).
Now we compute a curvature layer (Figure 5 bottom center), and them combine the curvature with the buffer layer, generating a new layer whose value is
the curvature on the buffer area (Figure 5 bottom right shows the result for one
fragment).
Finally the report of these layers gives the surface of each edge with every
curvature value.
6
Conclusions
In this paper we have presented an innovative approach to Cultural Heritage
Information Systems. This new strategy is based on the direct association of
layered data with the surface of an object, allowing the management, visualization and analysis of any kind of information (spatial or non-spatial data) related
to any cultural heritage artefact or archaeological site.
Following this approach, an application has been developed, which runs on
a personal computer. The application is being tested on real work at two emblematic cultural heritage sites in Andalusia: The Alhambra Palace in Granada
and the Roman city of Italica, close to Seville. The main goal of these tests is to
ascertain the effectiveness of the proposed system at a conceptual level.
Acknowledgements
This work has been founded by the Andalusian Science Ministry (Consejerı́a de
Innovación Ciencia y Empresa de la Junta de Andalucı́a) under grant PE09-TIC5276. All models are property of the Patronato de la Alhambra y del Generalife
and Museo de la Puebla de Don Fadrique.
References
[Diam10] T. Diamanti, P. Diarte Blasco, A. Guidazzoli, M. Sebastián López and E.
Toffalori ViSMan: an Open-Source Visualization Framework for Virtual Reconstructions and Data Management in Archaeology. The 11th International Symposium on
Virtual Reality, Archaeology and Cultural Heritage. VAST (2010)
[Feli11] A. Felicetti, F. Niccolucci: A Repository for Heterogeneous and Complex Digital Cultural Objects. The 12th International Symposium on Virtual Reality, Archaeology and Cultural Heritage. VAST (2011)
[Kata09] M. Katsianis, S. Tsipidis, K. Kotsakis, Alexandra Kousoulakou: A 3D digital
workflow for archaeological intra-site research using GIS. Journal of Archaeological
Science 35 (2008)
[Nete08] M. Neteler, H. Mitasova: Open Source GIS: A GRASS GIS Approach.
Springer, New York 2008
[Pena11] S. Pena Serna, R. Scopigno, M. Doerr, M. Theodoridou, C. Georgis, F. Ponchio, A. Stork: 3D-centered media linking and semantic enrichment through integrated searching, browsing, viewing and annotating. The 12th International Symposium on Virtual Reality, Archaeology and Cultural Heritage. VAST (2011)