Proceedings 18 th International Symposium CIPA 2001 
Potsdam (Germany), September 18 - 21, 2001 
THE APPLICATION OF REVERSE ENGINEERING TECHNOLOGY 
IN MODEL RE-CONSTRUCTION FOR CONSERVATION 
Prof. dr. ir. Bert Lauwers, Dr. ir. Qiongyan Li 
Katholieke Universiteit Leuven - Division PMA 
Celestijnenlaan 300B, B-3001 Heverlee, Belgium 
Phone: +32-16-322480, E-mail: bert.lauwers@mech.kuleuven.ac.be 
KEY WORDS: Conservation, Reverse Engineering, Modelling 
ABSTRACT 
This paper describes the development of a prototype software for the textured surface re-construction of buildings starting from depth 
maps derived from camera images. A depth map model represents the image information (x,y,z, texture) in an absolute working co 
ordinate system. The surface re-construction consists then of three steps: 3D-point treatment (3D point cloud processing), meshing 
and texturing. 
In a first step, the 3D point cloud processing module merges different depth maps together. The result is a non-optimised point cloud, 
which often contains too many points, certainly in the overlapping zones of two or more depth maps. This huge amount of data 
reduces the processing time afterwards and is often not needed for further modelling purposes. Therefore, the 3D point cloud 
processing module uses different filtering algorithms to reduce the number of points. 
In a second step, the point cloud is triangulated to a mesh. Advanced meshing as well as smoothing and noise reduction algorithms 
have been developed in order to increase the appearance of the generated mesh. The implementation of the point cloud processing 
and meshing algorithms module is based on existing technology (software) used for reverse engineering applications in mechanical 
engineering. Reverse engineering means the measuring of wooden or clay models succeeded by the reconstruction of the geometrical 
CAD model. 
In a third step, the mesh is textured. The texture of the meshed model is defined based on the original images. Specific algorithms 
have been developed to define the best image in case the triangle vertices do not reference to the same image or to multiple images. 
1. INTRODUCTION 
Many people involved in conservation still work in a rather traditional way. They use hand-measured (tapes, plumb-bobs, levels,...) 
or instrument based (theodolite, total station) survey methods. At the same time, more and more architects are shifting towards 
computer-aided design for new buildings and they also try to apply these programs to renovation or conservation projects. However, 
the number of tools available to accomplish the task of “getting the existing building into the CAD program’ is limited, and mainly 
directed to “translate” traditional methods to CAD. Based on a limited number of actually measured points, 2D plans, sections or a 
3D-model, are constructed. Mainly due to the difficulties of drawing irregular shapes in CAD programs, this also often results in 
‘idealised’ representations of buildings. As an example, Figure 1 shows an idealised façade drawing of an, in reality much more 
deformed, timber frame. 
a 
V 
8 
Figure 1: Idealised CAD representation (left) of a historical building (right) 
Apart from programs that only help to import measurement data, applications for digital photogrammetry are developed. They do not 
require very special equipment and allow deriving accurate 3D measurements from photographs inside a CAD program. These 
programs [Van Baien 95, Nuyts 2000] enable the user to measure a higher density of points in 3D and construct a drawing or model. 
However, the user is still confronted with the CAD program’s restrictions, especially working in 3D on non-orthogonal entities is 
very difficult.