El-Hakim, Sabry
A PRACTICAL APPROACH TO CREATING PRECISE AND DETAILED 3D MODELS FROM
SINGLE AND MULTIPLE VIEWS
Sabry EL-HAKIM
Visual Information Technology (VIT) Group
Institute For Information Technology, National Research Council Canada
Sabry.El-Hakim@nrc.ca
Working Group V/3
KEY WORDS: Image registration, Modeling, Multi-sensor, Scanner, Segmentation, Surface Reconstruction, VR
ABSTRACT
This paper describes a general system for creating geometrically correct and complete 3D models of objects and
environments. The main process uses digital images and applies rigorous photogrammetric techniques. Once the images
are registered using bundle adjustment, points in selected locations are measured in all the images where they appear.
However, there are usually many parts of the scene that appear in only one image due to occlusions or lack of features.
These parts can be reconstructed in the 3D space from the coordinates in a single image and the mathematical model of
the surface determined by fitting a function to existing surface points. This produces a sampled geometry in the form of
points in the three-dimensional space however the connectivity or the topology is not known and must be determined
someway. The proposed approach relies on interactive point segmentation and automatic triangulation. Counting on
images for modeling is limited because the features that can be extracted are usually fewer then the required level of
details. To overcome this problem two options are implemented. First, large number of points can be automatically
added to surfaces of known shape, such as spheres, cylinders and quadrics, using a polygon subdivision method. The
second option integrates data from range sensors that can densely digitize the surface where filled out details are
required. The presented approach will be assessed by several examples representing various types of objects and sites.
1 INTRODUCTION
The computer reconstruction of objects and environments has a wide range of applications. Manufacturing and virtual
prototyping [Gomes de Sa et al], reverse engineering [Thompson et al, 1999], urban design and analysis [Teller, 1998],
architecture [Liebowitz et al, 1999], and cultural heritage [Sablatnig et al, 1996], are just few examples. Methods to
acquire the 3D data and reconstruct shapes have been progressing rapidly in recent years. However, many problems
remain to be solved and no approach is suited for all applications and all types of object and environment. This paper
presents an approach that aims to be limitless. It is semi-automated, can generate 3D data from multiple or single
images, can integrate data from active range sensors, and can automatically add points on surfaces of known shapes.
The resulting models are accurate, complete, easily created, and not restricted to specific application.
In this introduction the state of the technology is summarized followed by an overall description of our approach. In the
second part of the paper details of the approach are given. In the third part an evaluation of the geometric accuracy and
the ability to model details using various types of data is presented. Finally, some conclusions and future research
directions are outlined.
1.1 The Current 3D Digitizing and Modeling Techniques
Techniques for 3D digitizing and modeling have been rapidly advancing over the past few years. The ability to capture
details and the degree of automation vary widely from one approach to another. One can safely say that there is no
single approach that works for all types of environment and at the same time is fully automated and satisfies the
requirements of every application.
The process of creating 3D models from real scenes has a few well-known steps [figure 1]: data collection, data
registration, and modeling (geometry, texture, and lighting). There are many variations within each step, some of which
are listed in figure 1. Approaches that skip the geometric modeling step also exist. For example, panoramas [Szeliski et
al, 1997 and Chen, 1995], light fields [Levoy et al, 1996], and image-based rendering (images plus depth) [Kang, 1999],
are popular for applications that require only visualization or limited walkthrough. However, the lack of geometric
model impedes the accuracy and the freedom to render the environment from arbitrary viewpoints.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000. 203