Structure and motion from long image sequences

When humans are presented with a sequence of monocular views of a rotating object, they have a vivid perception of the three-dimensional structure of the object. In addition, if an object in the field of view is moving, humans can more easily separate the object from the background than if it were stationary. This thesis presents a computer algorithm which, given a sequence of digital intensity images of moving objects, will separate the images into regions showing distinct objects, and for those objects which are rotating, will calculate the three-dimensional structure and motion. The method involves two major steps: finding and tracking feature points on the objects in the images, and determining the structure and motion of the objects from these feature point trajectories. To find the trajectories, the features of interest in the input images are enhanced, and instead of detecting the features in a single frame, the algorithm then detects the paths of the features as they move over several frames. To determine the structure and motion of the objects, the trajectories must be partitioned into groups corresponding to the different objects, and then for each of these groups, the three-dimensional structure and motion must be found. The structure and motion estimation algorithm requires that the objects are rigid, that the rotational motion is constant, and that the scene containing the objects is orthographically projected on the image plane. This work describes experiments using synthetic data to test the noise sensitivity of the structure and motion estimation algorithm under a variety of conditions. In addition, two real image sequences are used to test the complete algorithm, from tracking the features to finding the structure and motion of each object.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Electrical Engineering

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2011

Author dc:creator

• Debrunner, Christian Hans

Contributors dc:contributor

• Ahuja, Narendra

• Engineering, Electronics and Electrical
• Computer Science

Statement dc:rights

• Copyright 1990 Debrunner, Christian Hans

Language dc:language

eng

Identifier

AAI9114219
(UMI)AAI9114219

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/20949