Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Dynamic Vision Shape from Motion

Casagrande, Erik (2003) In MSc Theses
Department of Automatic Control
Abstract
Dynamic vision is a class of problems in computer vision that calculates properties of the3D world through the dynamics of the system observing by cameras. In this thesis we introduce the issue of motion and shape estimation with the aid of a single camera and for particular objects as points, planar surfaces and polyhedrons. This problem is well known in literature as Shape from Motion. We are interested in building a framework that, starting from a sequence of images that has been recorded from a camera, achieves the geometrical and dynamics properties of the 3D scene. In the first part we inspect the motion estimation through the 2D frames recorded by the camera. In particular we present a possible recursive approach for Optical Flow... (More)
Dynamic vision is a class of problems in computer vision that calculates properties of the3D world through the dynamics of the system observing by cameras. In this thesis we introduce the issue of motion and shape estimation with the aid of a single camera and for particular objects as points, planar surfaces and polyhedrons. This problem is well known in literature as Shape from Motion. We are interested in building a framework that, starting from a sequence of images that has been recorded from a camera, achieves the geometrical and dynamics properties of the 3D scene. In the first part we inspect the motion estimation through the 2D frames recorded by the camera. In particular we present a possible recursive approach for Optical Flow estimation. The second part introduces a new class of problems in system theory that are suitable for applications of computer vision. This subject has been called perspective system theory. We consider typical issue as observability, identifiability and realization theory for this branch of problems. Results has been shown through use of algebraic and recursive estimation algorithms. (Less)
Please use this url to cite or link to this publication:
author
Casagrande, Erik
supervisor
organization
year
type
H3 - Professional qualifications (4 Years - )
subject
publication/series
MSc Theses
report number
TFRT-5707
ISSN
0280-5316
language
English
id
8848076
date added to LUP
2016-03-19 17:36:41
date last changed
2016-03-19 17:36:41
@misc{8848076,
  abstract     = {Dynamic vision is a class of problems in computer vision that calculates properties of the3D world through the dynamics of the system observing by cameras. In this thesis we introduce the issue of motion and shape estimation with the aid of a single camera and for particular objects as points, planar surfaces and polyhedrons. This problem is well known in literature as Shape from Motion. We are interested in building a framework that, starting from a sequence of images that has been recorded from a camera, achieves the geometrical and dynamics properties of the 3D scene. In the first part we inspect the motion estimation through the 2D frames recorded by the camera. In particular we present a possible recursive approach for Optical Flow estimation. The second part introduces a new class of problems in system theory that are suitable for applications of computer vision. This subject has been called perspective system theory. We consider typical issue as observability, identifiability and realization theory for this branch of problems. Results has been shown through use of algebraic and recursive estimation algorithms.},
  author       = {Casagrande, Erik},
  issn         = {0280-5316},
  language     = {eng},
  note         = {Student Paper},
  series       = {MSc Theses},
  title        = {Dynamic Vision Shape from Motion},
  year         = {2003},
}