CIPA 2005 XX International Symposium, 26 September - 01 October, 2005, Torino, Italy 
802 
MULTISCALE THREE-DIMENSIONAL SURVEYING FOR CONSERVATION TASKS: 
A PILOT CASE FOR THE FUSION OF RANGE-SCANNING ON ARCHAEOLOGICAL 
SITES 
J.J.Femández-Martin 3 J.I.SanJosé 3 , M. Gonzalo b , J. Martinez 3 and J.Finat b 
3 LFA-DAVAP, ETS Arquitectura, Univ. de Valladolid, 47014 Valladolid, Spain 
b MoBiVA-DAVAP Group, Lab. 2.2, R+D Building, M.Delibes Campus, 47011 Valladolid, Spa, 
juanjo@ega.uva.es, jesussanjose@usuarios.retecal.es, marga@infor.uva.es, jmr@ega.uva.es, jfinat@agt.uva.es. 
KEY WORDS: Close range digital Photogrammetry, 3D Laser scanning, Cultural Heritage, Computer Graphics. 
ABSTRACT 
Integration of 3D data in digital photogrammetry is a challenge in solid object modelling and visualization of large scale scenes. 
Geometric primitives are need for a robust management of large reduced scenes, and an accurate visualization based on dense 
information is required for specific complex regions. Two complementary strategies for the fusion of high resolution digital 
photogrammetry and laser scan are developed in this paper. The first one is oriented towards the generation of a 3D geometric model 
based on a dense depth map linked to a large number of images. The second one is oriented towards the management of the whole 
3D scene with an intelligent decreasing of the file size adapted to expected primitives. Our contribution is linked to the automatic 
identification of geometric primitives, adaptation of warping techniques and the development of a voxelization with two levels of 
detail. On the way, we develop some interfaces for the management of 3D information from CAD models and the connection 
between unordered clouds of points and G1S data, relative to the environment with common geometric CAD primitives supported on 
2D and 3D information. We illustrate this hybrid approach with the surveying of a Spanish medieval castle (Trigueros del Valle, 
Valladolid) and a small chapel of the first gothic style in the monastère of Santa Maria de Valbuena (Valladolid), including a fusion 
of discrete clouds of points captured with triangulation and time-of-flight 3d lasers. 
1. INTRODUCTION 
The fusion of information in a common 3d framework is a 
challenge for the application of information technologies in 
surveying Cultural Heritage. The combination of topographic 
networks, ortophotos and laser-scans poses methodological and 
functional problems for planning, capturing and processing 
information to achieve a surveying with professional 
requirements. Inputs from laser scans are given by unordered 
clouds of 3d points, whereas inputs from high resolution views 
are given by 2d pixels on the camera plane. High resolution 
image-based approaches are focused towards matching 2d 
views by means of some reconstruction method. In digital 
photogrammetry, image-based approaches use colour properties 
for surfaces and bundles adjustment for geometric primitives; 
computer vision provides software tools for assisting the 
reconstmction of 3D scenes [Hartley and Zisserman, 2000]. 
Fusion of image-ranging approaches to Archaeological 
surveying has received attention in a high number of recent 
works [Guidi et al 2003]. The adaptation of dense 
reconstruction methodology to large datasets arising from laser 
scans is an on-going research. In our case we work directly on 
dense clouds arising from very large laser scanning files, with 
software tools for superimposing geometric structures (meshes, 
textures, PL or PS-surfaces) to clouds of points. 
Surveying and visualization of large urban environments or 
complex archaeological sites are typical problems where a 
multirresolution approach is required; a large number of 
contributions can be found in recent proceedings of CIPA 
Symposia, and ISPRS Workshops, focused towards surveying 
of Architectural and Archaeological Cultural Heritage. Large 
scale GIS applications use geometric meshing for multiscale 
rendering. Triangular Irregular Networks (TINs) provide a 
flexible solution for multirresolution depending on the chosen 
scale reduction [Berg et al, 2000]. However, large datasets 
arising from laser scans requires the development of processing 
tools for subdivision and re-grouping in 3d modelling. Both 
processes modify substantially the superimposed meshed 
structures. Smart meshing reduction depends on a hierarchy of 
primitives from the coarsest level (topological, geometric) to the 
finest one (architectural, topographic). On the other hand, direct 
CAD modelling of urban environments is benefited from the 
accurate and dense information capture arising from laser scans 
[Sanjose et al, 2005]. The combination of global flexibility of 
textured models with local accuracy of geometric primitives 
requires the development of specific processing and 
visualization software tools for satisfying flexibility of TINs 
with robustness/accuracy of geometric primitives. To satisfy 
both requirements, it is necessary to develop more flexible 
software tools for 1) a coherent interpolation between identified 
primitives, and 2) a compatible adaptability to adjacency 
conditions for GIS applications. 
There are mid-level computer vision software tools for the 
scene interpretation (perspective models) in terms of piecewise- 
1 inear or piecewise-smooth superimposed structures (normal 
vectors to triangles, contours extraction, correspondingly). The 
interpretation in terms of mid-level primitives requires a post 
processing and a high human/computational cost. It is 
commonly acknowledged that it would be desirable to obtain 
such information from dense maps, in a more direct way. Some 
work has been performed along last years for obtaining 3D 
information from dense depth maps. Information management 
by means of CAD software tools requires the design and 
implementation of algorithms for subdividing clouds of points 
according to their volumetric significance. In this work, we 
develop an approach including volumetric subdivisions based in 
grouping of tetrahedral decompositions fulfilling conditions 
similar to Delaunay. Computer Graphics tools have been 
applied for rendering large fortified environments from high 
resolution views [Bacigalupo and Cesari, 2003]. Currently, we 
develop interfaces for the management of 3d information with 
different levels of detail, depending on the urban surveying or 
archaeological visualization requirements for the insertion of 
Computer Graphics tools in large scanned environments. 
To achieve this goal, we have developed a hybrid approach with 
two resolution levels which are labelled as coarse and fine,