Automated Construction of Environment Models by a Mobile Robot

Only available on StudyMode
  • Topic: 3D scanner, Robotics, Voronoi diagram
  • Pages : 41 (12216 words )
  • Download(s) : 17
  • Published : February 12, 2013
Open Document
Text Preview
THESIS PROPOSAL
Automated Construction of Environment Models by
a Mobile Robot
Paul Blaer
Computer Science Department, Columbia University
December 9, 2004

Contents
1 Introduction

1

2 Related Work

3

2.1

Model Based Methods . . . . . . . . . . . . . . . . . . . . . . . . . .

4

2.2

Non-Model Based Methods . . . . . . . . . . . . . . . . . . . . . . . .

5

2.2.1

Volumetric Methods . . . . . . . . . . . . . . . . . . . . . . .

5

2.2.2

Surface-Based Methods . . . . . . . . . . . . . . . . . . . . . .

6

View Planning for Mobile Robots . . . . . . . . . . . . . . . . . . . .

7

2.3

3 Overview of Our Method

4 Initial Modeling Stage

8

11

4.1

Planning the Initial Views . . . . . . . . . . . . . . . . . . . . . . . .

11

4.2

Moving to the Chosen Scanning Locations . . . . . . . . . . . . . . .

18

4.2.1

Path Planning . . . . . . . . . . . . . . . . . . . . . . . . . . .

18

4.2.2

Computing the Generalized Voronoi Diagram . . . . . . . . .

19

4.2.3

Computing the Path . . . . . . . . . . . . . . . . . . . . . . .

22

4.2.4

Localization and Navigation . . . . . . . . . . . . . . . . . . .

23

Scan Acquisition and Integration . . . . . . . . . . . . . . . . . . . .

24

4.3

5 Final Modeling Phase

25

6 Road Map to the Thesis

31

1

1

Introduction

Accurate three-dimensional models of large outdoor structures, such as buildings and their surroundings, have many uses. They can provide an educational walk around a structure that is thousands of miles away. These models can allow engineers to analyze the stability of a structure and then test possible corrections without endangering the original. They allow us to preserve historical sites that are in danger of destruction, and they allow us to preserve archaeological sites at various stages of an excavation. Construction of such models is particularly useful when the original blue prints for the structures are either not available or inaccurate. In many cases; structures, ranging from third world oil refineries to medieval cathedrals, were developed over time and from many different designs, so that there is no longer one true blue print. In all of these cases, it is useful to have an accurate computer based 3-D model of the large scale outdoor structure. Methods for acquiring such models have progressively increased in accuracy and have evolved from manual methods to more automated methods. At the simpler end of the spectrum, one can send a team of surveyors with a theodolite to the site in order to take measurements of the structure and then have a designer put together a model from those measurements. Such a model might look relatively accurate and, in the case of its larger features, be geometrically accurate. However, it does not tell the whole story. It would be extremely inefficient to hand survey all of the small features and details on the surface of the structure. These features would likely not have been surveyed, and the model would be based mostly on the designer’s best guess.

More sophisticated tools do exist. There are a number of laser range scanners on the market that will sweep a beam across a large portion of the structure and return a dense point cloud of measurements. Armed with these more sophisticated instruments, one can take a number of scans around the structure and fuse them into a single point cloud that accurately represents the structure. The point cloud could then be triangulated into a mesh to give an accurate model. With a sufficient density of the point cloud that the scanner returns, one can generate models that are accurate to a centimeter or better. Now, instead of having a simplistic model of the structure, we have an accurate model appropriate for preservation and analysis. Although the models are now far more accurate and the acquisition process is faster and more automated, there is still a major human...
tracking img