|
" Control-Based Reconstruction and Pose-Estimation "
Islam, Bipul
Sandhu, Romeil
| Document Type
|
:
|
Latin Dissertation
|
| Language of Document
|
:
|
English
|
| Record Number
|
:
|
1110689
|
| Doc. No
|
:
|
TLpq2488002177
|
| Main Entry
|
:
|
Islam, Bipul
|
|
|
:
|
Sandhu, Romeil
|
| Title & Author
|
:
|
Control-Based Reconstruction and Pose-Estimation\ Islam, BipulSandhu, Romeil
|
| College
|
:
|
State University of New York at Stony Brook
|
| Date
|
:
|
2020
|
| student score
|
:
|
2020
|
| Degree
|
:
|
Ph.D.
|
| Page No
|
:
|
150
|
| Abstract
|
:
|
One of the pertinent problems in robotic vision is scene understanding from data-acquired by a network of 2D imaging sensors. In particular, two of the fundamental problems in vision in this regard are shape and pose estimation of scene-objects. Most modern learning/estimation methods solve these problems by formulating them as optimization problems that try to model the data-space into a tractable form. In real life applications, due to the dynamic nature of real-world scenarios and the dynamic nature of data inconsistencies, such formulations are often incomplete. That is, there always will exist an information gap in any given modeling effort especially under difficult imaging conditions found in aerospace, medical, and space applications. This dissertation is to study the aforementioned key problems in robotic vision in a variational and control setting for which we devise strategies to manage algorithm performance degeneration. From this, we then reinterpret the reconstruction problem as a multi-agent network problem in which we are interested in understanding its dynamical properties; i.e. how to potentially understand statistics over a time-varying networks. Nevertheless, the main goal is to provide a control variational framework that is capable of handling dynamic real-world data-inconsistencies from a geometric perspective. Ultimately, such advances can be applied to a general body of problems not limited to robotic vision.
|
| Subject
|
:
|
Applied mathematics
|
|
|
:
|
Artificial intelligence
|
|
|
:
|
Bioinformatics
|
|
|
:
|
Computer science
|
|
|
:
|
Robotics
|
| |