Huihai Lu
Title of thesis Object recognition and tracking
Position: Student
Degree registered: PhD
Duration of study (October 2002 – Now)
Financial support BTexat
Supervisor Prof. Mohammed Ghanbari and Prof. Ian Henning
Current affiliation hlu@essex.ac.uk
Abstract
This work isolates and tracks non-homogeneous semantic objects by employing
region based analysis in a Binary Partition Tree (BPT). In addition to colour, texture
and shape information, we propose that the relational information plays an important
role in identifying complex objects. Target characteristics are captured in a bank of
user defined models. Matching the tree with these models is formulated under a
Bayesian framework. In the case of tracking, the Multiple Hypothesis Tracking
(MHT) algorithm is also incorporated into this Bayesian framework giving a unified
view of the problem. The final algorithm has the ability of simultaneous object
detection and multiple object objects tracking.
Publications
E. L. Andrade, J. C. Woods, H. Lu, I. Henning and M. Ghanbari, “Generic object
registration using multiple hypotheses testing in partition trees,” European workshop
on the integration of knowledge, semantics and digital media technology, Nov. 2004,
London, UK.