Jigital 
di Milano, 
). 
one 
pt GIS for 
tional 
  
AUTOMATIC REGISTRATION OF 3-D VIEWS 
R. Bologna“, A. Guarnieri°, M. Minchilli", A. Vettore“ 
“ Dept. of Architecture and Town-Planning — Polytechnic of Bari (Italy) 
? Dept. of Architecture and Planning — University of Sassari (Italy) 
* CIRGEO (Interdept. Research Center of Cartography, Photogrammetry, Remote Sensing and GIS) 
University of Padua — 35020 Legnaro (PD) — AGRIPOLIS - Italy 
e-mail: cirgeo@unipd.it ; antonio.vettore@unipd.it 
Commission V, WG V/4 
KEY WORDS: 3D models, automatic registration, range data 
ABSTRACT 
Building of 3D models for single objects so as for whole environments, represents a continuosly and quickly evolving new research 
field of the computer science. The arising interest in this field is strictly related with the increasing availability of more powerful 
CPUs, overall in terms of PCs, which made possible, for the first time, to effectively manage complex 3D models off the research 
centers, as well. A new set of issues has to be addressed in the 3D modeling of real objects. A lot of data are needed about the object 
surface or volume, which have then to be aggregated, regardless the data format and the acquisition device used, in order to get the 
final model. Actually, the data registration step requires a human operator, which were able to provide a first rough alignement 
between acquired data. This approach is often time-consuming, increases the final cost of the 3D model and represents the major 
limit to the wide spreading of real object models. Alternatively more sofisticated range data acquisition devices can be used, such a 
range sensor mounted on a robotic arm with six degree of freedom, but anyway it is a very expensive modeling system. 
In the light of topics previously exposed, a fully automatic range data registration system has been developed. This system is able to 
execute all the steps needed for 3D modeling of real objects in automatic way or at least minimizing as more as possible the human 
intervention, without any other information but the range data only. 
In this paper, the subsystem for alignement of range data pairs is presented. The work draws the idea from A. E. Johnson[1], which 
proposed an innovative solution for the recognition of similarities between 3D surfaces, introducing the spin-image concept. The 
advantage of this approach rely on high computational robustness and effectiveness, which allows to employ standard market-level 
CPUs. On the ground of the spin-image concept, a full data registration system was developed, in which the overlapping areas of two 
adjacent data set are automatically recognized, thus allowing to correctly align the two whole data sets. 
1. INTRODUCTION 
Building of 3D models for single objects so as for whole 
environments, represents a continuosly and quickly evolving 
new research field of the computer science. The arising interest 
in this field is strictly related with the increasing availability of 
more powerful CPUs, overall in terms of PCs, which made 
possible, for the first time, to effectively manage complex 3D 
models off the research centers, as well. In the wide area of 3D 
models we can distinguish two main categories: CAD or real 
object models. 
The models belonging to the first class are realized with a 
workstation and are fully under the responsability of the 
designer. These models can be created in a CAD environment to 
build real objects, such for instance mechanical pieces, and 
often they are then translated in the VRML format in order to be 
distributed on the web. Sometimes an inverse approach is 
needed: building a 3D model from a real object. Motivations at 
the ground of this choice are different, i.e. such a models can be 
used for reverse engineering or also in the medical field, to get 
more qualitative and quantitative information than common 2D 
images could offer. In this last case it should be considered the 
3D modeling of anatomical parts such theeth or bones. Another 
application of the second class of 3D models deals with the 
need of filling an virtual environment with real objects, in order 
to get a faithful copy of a real environment, such as the inside of 
a museum or historical building. 
Furthermore, 3D models of both abovementio-ned classes 
represent an interesting tool for documentation and interactive 
visualization purposes. For example they allow to represent 3D 
objects more adequately than through single picture or 
collection of pictures, and to change easily the user point of 
view, providing in this way a useful tool by which the human 
direct inspection can be well simulated. 
A new set of issues has to be addressed in the 3D modeling of 
real objects. A lot of data are needed about the object surface or 
volume, which have then to be aggregated, regardless the data 
format and the acquisition device used, in order to get the final 
model. Actually, 3D modeling of real free-form surfaces consist 
of the following steps: 
1. Pairwise alignment of the 3D images; 
2. Global alignment; 
3. Fusion of the 3D data originally captured as clouds of points 
into 3D surfaces; 
4. Editing of possible surface holes due to minor missing data. 
Step 2) and 3) are already performed automatically, while step 
1) at the present requires a human operator, which were able to 
provide a first rough alignement between acquired data. This 
approach is often time-consuming, increases the final cost of the 
3D model and represents the major limit to the wide spreading 
of real object models. 
Step 4), nowadays performed manually, may not be necessary if 
an adequate amount of data is captured (which however may not 
always be feasible). 
—315—