RI | Thesis Oral | 29 Jul 2002

Robotics Institute Thesis Oral 29 Jul 2002
Place and Time | Abstract | Further Details | Thesis Committee

Robust Tracking and Structure from Motion with Sampling Method

Peng Chang
Robotics Institute
Carnegie Mellon University

Place and Time

NSH 3305
10:30 AM

Abstract

Robust tracking and structure from motion (SFM) are fundamental problems in computer vision that have important applications for robot visual navigation and other computer vision tasks. Although the geometry of SFM problem is well understood and effective optimization algorithms have been proposed, SFM is still difficult to apply in practice. The reason is twofold. First, finding correspondences, or "data association", is still a challenging problem, especially under degenerate conditions, such as occlusion and abrupt camera motion. Second, the result of SFM is often observed to be unreliable under uncontrolled environments, especially when the system is undergoing unexpected dynamics.

This thesis aims to tackle both problems simultaneously. We attribute the difficulty of applying SFM in practice to the unmodeled noise usually seen in image data in uncontrolled environments. And we propose to integrate the SFM with tracking so that the tracking can reason about occlusion and unexpected camera motion and become robust against those extreme conditions by properly detecting them. Representing the uncertainty is of the most important aspect of this work so that a probabilistic framework can be applied. We propose a sampling method to capture the uncertainty in both tracking and SFM. The uncertainty is naturally represented with sample sets. A probabilistic filtering algorithm is developed to propagate the uncertainty through time. With the sample-based representation, our system can capture the uncertainty in tracking and SFM under degenerate conditions, such as occlusion and extreme camera motion, therefore exhibiting improved robustness in those degenerated conditions, which are usually difficult for the traditional approaches. We believe that with this sampling-based probabilistic framework, we are one step closer to an SFM system that can perform reliably in real robot navigation tasks.

Further Details

A copy of the thesis oral document can be found at http://www.cs.cmu.edu/~peng/thesis.pdf.

Thesis Committee

• Martial Hebert, Chair
• Robert Collins
• Jianbo Shi
• Michael Black, Brown University

This page maintained by robotwebmaster@ri.cmu.edu.