Clifford, William
(2023)
Multiple View Texture Mapping: A Rendering Approach Designed for Driving Simulation.
PhD thesis, National University of Ireland Maynooth.
Abstract
Simulation provides a safe and controlled environment ideal for human
testing [49, 142, 120]. Simulation of real environments has reached
new heights in terms of photo-realism. Often, a team of professional
graphical artists would have to be hired to compete with modern commercial
simulators. Meanwhile, machine vision methods are currently
being developed that attempt to automatically provide geometrically
consistent and photo-realistic 3D models of real scenes [189, 139, 115,
19, 140, 111, 132]. Often the only requirement is a set of images of
that scene. A road engineer wishing to simulate the environment of a
real road for driving experiments could potentially use these tools.
This thesis develops a driving simulator that uses machine vision
methods to reconstruct a real road automatically. A computer graphics
method called projective texture mapping is applied to enhance
the photo-realism of the 3D models[144, 43]. This essentially creates
a virtual projector in the 3D environment to automatically assign image
coordinates to a 3D model. These principles are demonstrated
using custom shaders developed for an OpenGL rendering pipeline.
Projective texture mapping presents a list of challenges to overcome,
these include reverse projection and projection onto surfaces not immediately
in front of the projector [53]. A significant challenge was
the removal of dynamic foreground objects. 3D reconstruction systems
create 3D models based on static objects captured in images.
Dynamic objects are rarely reconstructed. Projective texture mapping
of images, including these dynamic objects, can result in visual
artefacts. A workflow is developed to resolve this, resulting in videos
and 3D reconstructions of streets with no moving vehicles on the scene.
The final simulator using 3D reconstruction and projective texture
mapping is then developed. The rendering camera had a motion
model introduced to enable human interaction. The final system is
presented, experimentally tested, and future potential works are discussed.
Item Type: |
Thesis
(PhD)
|
Keywords: |
Multiple View Texture Mapping; Rendering Approach; Driving Simulation; |
Academic Unit: |
Faculty of Science and Engineering > Computer Science |
Item ID: |
17831 |
Depositing User: |
IR eTheses
|
Date Deposited: |
14 Nov 2023 14:35 |
URI: |
|
Use Licence: |
This item is available under a Creative Commons Attribution Non Commercial Share Alike Licence (CC BY-NC-SA). Details of this licence are available
here |
Repository Staff Only(login required)
|
Item control page |
Downloads per month over past year
Origin of downloads