All Welcome

===============================================================

Principles of Industrial Image Processing
Dr Bernd Sommer, STVision GmbH
STVision GmbH
Bergstrasse 5a
85410, Haag / Germany

Abstract – Target of industrial inspection is the introduction of six-sigma error-free output in the production line. We give the specific requirements to image processing for this target. State of the art camera technology is outlined, as well as specific illumination systems, handling of components and efficient software technology. This includes vectorisation of algorithms, parallel processing, machine language optimization of core algorithms, and much more. Finally, a number of applications is outlined.

Speaker:

Dr Bernd Sommer is the Founder and CEO of STVision GmbH in Munich, Germany. He received his PhD in theoretical physics in 1977 from the University of Bochum, Germany, with a dissertation on electro-disintegration of deuterium at back scattering angles. After 2 years at a postdoc at Bochum, he joined Kontron Electronics in Munich as an R&D manager for vision systems and CAD/CAM system design. In 1989, Dr Sommer established a company for industrial image processing, initially with applications for die and wire bond inspection in semiconductor manufacturing. Since then, STVision has operated from Munich and from Neuchatel, Switzerland, with a diverse range of image processing applications.

All Welcome

===============================================================

Unintended Uses for a Speech Recognition System: Speaker Characteristics

Dr Michael Norris
Programmer, Software Engineering
Faculty of Information Sciences & Engineering
University oF canberra

Abstract – A speech recognition system aims to determine the spoken text from a continuous stream of audio. Speech sounds are matched against a population of statistical models of the acoustic properties of speech sounds. Such acoustic models are created by training on a large corpus of annotated speech. The acoustic models therefore depend on the population of speakers on which they were trained. Speech recognition systems include tools for adapting acoustic models to a particular speaker or population of speakers.

This study explores the possibility that by partitioning a corpus of speakers into groups having distinct characteristics and training acoustic models for each sub-population, statistics generated by the speech recogniser can be used to determine which sub-population the speaker is more likely to belong to. Therefore, parts of an existing speech recogniser can be used to determine characteristics of the speaker.

Experiments with determining speaker age and gender by this method will be presented and discussed.

Dr Michael Norris
Programmer
Software Engineering
Faculty of Information Sciences and Engineering
Univeristy of Canberra

Mathematical Modelling: Why, Where and How.

Geoff Mercer BSc PhD
Fellow, National Centre for Epidemiology and Population HealthAustralian National University, Canberra, ACT 0200, AUSTRALIA

Abstract – Mathematical modelling is a powerful tool when used wisely but is unfortunately often misunderstood and poorly implemented. Modelling is not a standalone discipline and crucial to the success of any modelling project is the communication with, and understanding of, everyone involved in the project. This is generally not something that is taught at either undergraduate or graduate level with most modelling courses concentrating exclusively on narrow mathematical or statistical aspects. These ideas will be discussed with a view to the type of undergraduate or graduate courses that could be offered. Three diverse examples of recent successful modelling projects will then be presented. Firstly, modelling is used to determine if receipt of the seasonal influenza vaccine was a risk for contracting pandemic influenza. Secondly, the behaviour of spray droplets impacting leaves is modelled for different adjuvants and leaf characteristics. Thirdly, optimal aerial maritime surveillance routes are determined for moving target ships when there are fuel and air crew constraints.

Dr Geoff Mercer BSc PhD
Fellow, National Centre for Epidemiology and Population HealthAustralian National University, Canberra, ACT, AUSTRALIA

The Uses of Mathematical Modelling

Dr Steve Barry

National Centre of Epidemiology and Population Health

ANU College of Medicine and Health Sciences

The Australian National University

Canberra, ACT, 0200 Australia

Abstract

The word “modelling” is used (and abused) in many contexts. I define it as follows: “Modelling”: simplifying a system to its essential elements to gain insight, understanding or predictive ability. There are many different types of ‘models’ – epidemiological animal models (ferrets stuck in cages giving each other the flu) through to systems models, computer simulations, statistical models, deterministic models and stochastic models. Modern ‘modelling’ now incorporates elements of all of these (even the ferrets) in order to find a useful and practical solution to the problem at hand.

This talk will use practical examples to give an overview of different modelling methods, and touch on some of my own work on drain placements on rural roads, maritime surveillance and invasive species spread. In particular I will talk about my work on the recent influenza pandemic and how modelling can be used to influence public policy.

I will be answering such questions as:

1) can border protection make a pandemic worse?

2) what is the optimal placement of drains on a rural road?

3) what is the optimal route in maritime surveyllance?

4) how long does it take to heat up a 2-tonne steel coil?

5) how can we model invasive species spread?

6) how to design a good breast pump.

7) what proportion of Salmonella comes from chicken?

The talk will be aimed for a general non-mathematical audience.

Speaker:

Dr Steve Barry did his undergraduate degree in applied mathematics at the University of Adelaide before doing a PhD at UNSW on how fluid flows through deformable porous materials. He applied this work to blood flow in capillaries while at Imperial College, London and to the design of breast pumps. He did post-doctoral work in the Dept. of Mechanical Engineering, University of Toronto, studying the ultrafiltration of biopolymers. He was a lecturer and senior lecturer at UNSW@ADFA where his research turned to industrial mathematics becoming involved in numerous diverse problems including crystal growth rates in cocoa butter, modelling spreading weeds and multilayer diffusion. His general research interests lie in the process of turning a real application into a mathematical model, and using this to find useful solutions. He is currently a Fellow at the National Centre of Epidemiology and Population Health working on modelling seasonal and pandemic influenza and source attribution in Salmonella. He has done modelling work for numerous organisations including Arnottes Biscuits, DSTO, Grape Research Corporation, GH Michel and Sons (wool scouring), Australian Government Department of Health and Aging, and NZ Steel.