LARGE-SCALE STRUCTURE-FROM-MOTION AND RELATED PROBLEMS: A MODERN SYNTHESIS

PhD Qualifying Examination


Title: "LARGE-SCALE STRUCTURE-FROM-MOTION AND RELATED PROBLEMS: A MODERN
SYNTHESIS"

by

Mr. Tianwei SHEN


Abstract:

This paper is a survey of theories and methods in large-scale 
Structure-from-Motion (SfM). Structure-from-Motion is a fundamental research 
topic that has been intensively studied in 3D computer vision. We will review 
this problem with a modern perspective. Over the years, SfM research has 
evolved from 2-view or 3-view minimal cases to a complex system that incorpo- 
rates knowledge from image retrieval, convex optimization, distributed 
computing and large- scale system designs. We will conduct the review by 
grouping references into several topics. Related topics include but not limited 
to efficient image collection matching, close-form linear minimal solutions of 
camera pose estimations, match graph constructions that are robust and 
resistant to ambiguous structures and noises, convex modelling of motion 
averaging and large- scale bundle adjustment in distributed settings. Classical 
methods and emerging approaches are both thoroughly investigated and discussed. 
At the end, we also point out several future directions from our own 
perspectives.


Date:			Wednesday, 31 August 2016

Time:                  	10:00am - 12:00noon

Venue:                  Room 3494
                         Lifts 25/26

Committee Members:	Prof. Long Quan (Supervisor)
 			Prof. Albert Chung (Chairperson)
 			Prof. Huamin Qu
 			Prof. Chiew-Lan Tai


**** ALL are Welcome ****