My Research

Machine Learning in Computer Vision

My group started work in the then new area of Machine Learning In Computer Vision in 1994, and the focus has been the development of machine learning and statistical analysis techniques for solving problems within computer vision. My research in this area utilises and extends powerful machine learning techniques and applies them to the complex problem of learning object models for obj ect recognition. Better recognition rates have been demonstrated using these tec hniques and a paper describing the techniques was presented at 16th Internationa l Conference on Machine Learning in 1999. Another paper describing a novel re-sampling technique for active vision systems and co-authored by myself, won the Best Student Paper at ICPR '98, the premier inte rnational conference in pattern recognition. Since then, my research has concentrated on new learning techniques for the e extraction of linear features in remotely sensed images, which present additi onal challenges such as new sources of noise and varying image resolutions and spectral characteristics. My journal article published in the International Journal of Photogrammetry and Remote Sensing on this topic, was the second most downloaded paper from the journal's online archives in both 2001 and 2002. Since 2005, my group has made significant contributions in the area of level set methods for segmentation and object extraction in images and also successfully moved machine learning techniques developed for vision into general data partitioning, with a paper presented at the European Conference on Machine Learning ECML 2007.

Since 2002, my group has been developing innovative learning techniques for medi cal images, including new incremental learning algorithms and innovative image a nalysis techniques for anatomy and feature extraction based on machine learning and knowledge acquisition. A prototype Computer Aided Diagnosis (CAD) system for automated diagnosis of diffuse lung diseases from CT images, integrating the ne w techniques has been developed. The medical imaging research was in collaborat ion with Medical Imaging Australasia as clinical and Philips Medical Systems Aus tralasia as industry partners. Since 2012, the medical imaging research has focussed on analysis of Magnetic Resonance Images, in paryicular multi-parametric MR analysis for prostate cancer recognition in collaboration with Prince of Wales Hospital, Sydney; and brain MR analysis and machine learning for detection of Mild Cognitive Impairment n collaboration with Centre for Healthy Brain Ageing, UNSW.

In the area of motion tracking and segmentation, my group is working on tracking and recognition techniques suitable for immersive environments, as well as face tracking. Experiments are conducted at the iCinema centre. We have developed active vision techniques for an autonomous and adaptive security camera and a human motion classification system for it. Activity recognition is an ongoing area of research. We have also developed algorithms and systems for high precision measurement of bacteria, touchless and touch-based interaction with bacteria biofilm videos, and an augmented reality system for remote viewing of and interaction with bacteria biofilm videos.

Software Engineering

My research in this area includes formal specification and verification of real-time reactive systems, and the techniques have been applied to robotic control design, real-time scheduling systems, design of adaptive reuse techniques for embedded systems and protocol verification.

My group has made significant contributions to the field of formal modelling and analysis of concurrent real-time systems. We proposed a formal verification methodology y for Statecharts based on a customised temporal logic, published in IEEE Transactions on Software Engineering. My research group proposed a formal framework for automated hardware component reuse, based on an innovative technique called forced simulation invented by us, along with practical algorithms that automatically generate device drivers to adapt off-the-shelf components to match the requirements. The results were published i n ACM Transactions on Design Automation of Electronic Systems 2001. A more efficient algorithm for forced simulation using tabled logic programming was then developed and published in IEE Proceedings Computers and Digital Techniques. My group also developed automated interface synthesis techniques for mismatched protocols in VLSI chip design within a synchronous modelling framework, presented in IEE Proc. Computers and Digital Techniques in 2005. My group recently extended this work to deal with on-chip protocol incompatibility in general. A highly expressive formal modelling language for state-based communication synchronous protocols was developed, which allows modelling of modern industrial protocols at a level close to Hardware Design Languages but with better analysis capabilities. An algorithm to check for protocol incompatibility, and generate a small `converter' (converter synthesis) if necessary, has been designed. We have also developed methods for design space exploration to allow for converter optimization, which are the first in the field. My group has studied synchronous and asynchronous models of time and verificatio n techniques for these using timed automata. A Simulink block for the synchronou s language Argos was built and presented. A tool to perform time trace inclusion checking between multiple system implementations mo delled as timed automata has been built and released to the research community.