TITLE: Machine Learning and Information Fusion in Computer Vision

PRESENTER: Xiongcai Peter Cai, http://www.cse.unsw.edu.au/~xcai/, xcai@cse.unsw.EDU.AU

AFFILIATION:School of Computer Science and Engineering, UNSW, http://www.cse.unsw.edu.au

DATE: Tuesday 14th March 2006

TIME: 14:00:00

PLACE: Level 4 Meeting room K17

ABSTRACT:

Machine learning and information fusion play an essential role in
computer vision due to their ability to handle uncertainty and multiple
cues. On the one hand, there exist a set of ill-posed problems in
computer vision, such as the stereo problem, which cannot be solved
without the help of multiple information sources. On the other hand,
model based vision has its strengths and weaknesses and may not be able
to manage all situations or solve a whole problem. This raises the need
for a learnable dynamic approach to handle different parts of the problem
or various situations where the system works.

A learning-based method for parameter tuning of object recognition systems
and its application to automatic road extraction from high resolution
remotely sensed (HRRS) images is presented. Our approach is based on region
growing using fast marching level set method (FMLSM), and machine learning
for automatically tuning its parameters. FMLSM is used to extract the shape
of objects in images. Parameters are introduced into the speed function of
the FMLSM to improve flexibility and reflect the variety of images. The
parameters are tuned using machine learning and utilizing background
knowledge. The primary contribution of our approach is the ability to learn
the parameters for a FMLSM model for object extraction.

This talk will describe our progress so far in detail and indicate the
future plan in this project.

BIOGRAPHY OF SPEAKER:

Xiongcai Peter Cai (Peter) is a PhD student at CSE. His research interest
includes Computer Vision, Machine Learning and Computer Graphics.

Host:

Arcot Sowmya

Seminar Convenor:

Van Hai Ho