- 62 - 
elements and for generating points in occluded or symmetrical 
parts is to do the work in the 3D space to find the new points 
then project them on the images using the known camera 
parameters. The main steps are shown in figure 3. We will now 
give some details on the use of seed points. 
A cylinder is constructed after its direction, radius, and position 
have been automatically determined from four seed points 
(figure 4-a) using quadric formulation [Zwillinger, 1996]. The 
ratio between the upper and the lower circle can be set in 
advance. It is set to less than 1.0 (about 0.85) to create a tapered 
column. From this information, points on the top and bottom 
circle of the column (figure 4-b) can be automatically generated 
in 3D resulting in a complete model. 
Figure 4. (a) Four seed points are extracted on the base 
and crown, (b) column points are added automatically. 
model, texture maps obtained from high-resolution color digital 
camera is necessary. Some scanners have a color camera 
attached to the scanner at a known configuration so that the 
acquired texture is always registered with the geometry. 
However, this approach may not provide the best results since 
the ideal conditions for taking the images may not coincide with 
those for scanning. Therefore, our approach allows taking the 
images at different time from scanning and at whatever 
locations that will be best for texture. Details of the texturing 
procedure are described in another paper [Beraldin et al, 2002]. 
4.3 Combining Image- and Range-Based Modeling 
First the model of the whole structure, except for the fine details 
and sculpted surfaces, is modeled using the image-based 
approach (section 4.1). The sections that require scanning will 
be modeled separately using the approach described in section 
4.2. Common points between the image-based model and the 
range-based models are used to register them in the same 
coordinate system. This is done interactively with our own 
software that can display and interact with images from various 
types of sensors and cameras. The next step is to automatically 
sample points from the range-based model along its perimeter 
and insert those into the image-based model. The triangulated 
mesh of the image-based model will be adjusted based on those 
new points so that when the range-based model is added into 
that region there will be no overlapping triangles. This will be 
shown in the case study in section 5. 
4.4 Landscape Visualization 
For windows and doors we need three (preferably four) comer 
points and one point on the main surface (figure 3). By fitting a 
plane to the comer points, and a plane parallel to it at the 
surface point, the complete window or door is reconstructed. 
For details on other elements, see [El-Hakim, 2002]. 
4.2 Range-Based Modeling 
The procedure for creating a triangular-mesh model from 3D 
images is summarized in figure 5. If the 3D data is presented as 
a set of registered images it is trivial to create a triangular mesh 
by simply triangulating each image. However, since there is 
often considerable overlap between the images from different 
views, a mesh created in this fashion will have many redundant 
faces. It is desirable to create a non-redundant mesh, in which 
there are no overlapping faces. The adopted technique has been 
developed over the years at our laboratory and Innovmetric 
Software Inc. [Soucy and Laurendeau, 1995] and has been 
implemented in Polyworks™ [Innovmetric, 2002]. 
Sensor Placement 
A 
Fill Gaps 
> Scanning 
Registration 
! 
Modeling 
Digital Images 
Texture Mapping < 1 
Figure 5: General procedure for range-based modeling 
Most laser scanners focus only on acquiring the geometry. They 
usually provide only a monochrome intensity values for each 
pixel as sensed by the laser. To acquire realistic look for the 
When images of the whole scene taken at large distances such 
as aerial images are available, panoramic images of the 
landscape can be created and integrated with the model of the 
structures. This shows the structures in their natural setting and 
increases the level of realism. A few joint points between the 
structures and the grounds are measured in 3D to be used to 
register the panorama with the structures. The procedure is 
similar to [Sequeira et al, 2001]. 
5. MODELING THE ABBEY OF POMPOSA 
The abbey of Pomposa near Ferrara, one of the most appealing 
Italian churches of the Romanesque period, is a complex made 
of several buildings that are part of one of the most important 
Benedectine monasteries. Founded in the seventh century, the 
Refectory, the Basilica, the Capitolary Hall and the Cloister 
form the core of the abbey. The bells tower was added in the 
eleventh century. The abbey is architecturally simple with 
planar stone surfaces. The façade is ornamented with several 
relief works of art carved in marble. There are also three arches 
decorated with brick and stonework. 
Details like the left wheel “rosone”, the peacock carvings on the 
left side, and one end column (figure 6) were scanned with the 
Bids 3D sensor [Beraldin et al, 1999]. The whole complex was 
imaged with an Olympus 4 mega-pixel digital camera. Figure 7 
shows the model of the front of the church with the main 
structural elements. Figure 8 shows a close up of the general 
model with added 8 new points from the trim of the wheel and 
the re-triangulated mesh. The hole shown in the model is where 
the model of the scanned wheel will fit. Figure 9 shows the 
detailed solid model, without texture, of the wheel and the 
peacock after being added to the main model. A close up on 
part of the middle of the wheel showing a detailed wire frame is 
given in figure 10. The textured model of the same section is