Machine vision – theme-day in the UK

Citation

(2001), "Machine vision – theme-day in the UK", Kybernetes, Vol. 30 No. 2. https://doi.org/10.1108/k.2001.06730bab.005

Machine vision – theme-day in the UK

Machine vision – theme-day in the UK

Keyword: Cybernetics

A theme day organised in the United Kingdom was concerned with "Research into vision and imaging technologies" and it examined their application in a range of industrial functions. Held in June 2000, it brought together the UK research community and provided an up-to-date view of both work in progress and future aspirations. It was a theme day organised by the UK's Engineering and Physical Sciences Research Council (EPSRC), and reported in the council's Impact, Issue 28, November 2000, p. 4. The report says that:

The good news is that, overall, machine vision research in the UK is in reasonably good health: "Intellectually, the research can hold its own with the rest of the world," says Dr Phil Greenway, Head of the Advanced Information Department at BAe Systems in Bristol, and one of the evaluators who artended EPSRC's Machine Vision and Image Processing Theme Day held in June at the Institute of Civil Engineers.

There are several centres of excellence and the UK is particularly strong in 3-D imaging (see examples of EPSRC funded projects below and on the next page).

Machine Vision – the use of devices for optical non-contact sensing to automatically receive and interpret an image of a real scene, in order to obtain information and, or, control machines and processes – covers a range of research areas.

The community has been effective at training postgraduate students and Research Assistants (RAs) who go on to a wide variety of careers in, for example, finance, the software industry, and academia, as well as a range of Vision companies (including start-up businesses).

However, I believe strongly that "systems engineering" is both an important and sadly neglected aspect of machine vision research, says Dr Greenway. Much of the work is driven by a bottom-up, let's-see-what-we-can-do-with-this-idea approach. So perhaps it isn't too surprising that the resulting algorithms and methods often seem to fail to make much impact on wider, real world problems.

Dr Greenway feels that what is required is a complementary top-down, let's-see-what-is-needed approach. Underpinning this should be a rigorous and well-principled approach to performance measurement.

This view is motivated by the industrial need to produce well engineered systems incorporating vision, he adds. To achieve this, I feel that the process of developing machine vision technologies needs to accelerate its move from educated guesswork to an engineering basis founded on quantitative measurement. In particular, such measurement is essential, possible, and useful.

The need to address performance assessment is now being recognised by the research community.

Dr Phil Greenway from BAe Systems believes that attention needs to be given to the exploitation of information from multiple sources among which vision may be only a minor contributor. He also said that the new IRC in medical image processing which Oxford University's BP Professor of Information Engineering, Mike Brady, presented at the theme day, is exactly the kind of approach likely to be successful in the future. He stated that:

As a concept of the way and scale of its funding, and approach it's taking in the research, Mike's IRC has proper critical mass teams, covering all the ground they need to cover to get a good handle on how to solve all of the problems.

Inevitably they'll end up solving a whole set of problems that are different to the ones they've set out to solve. That's the nature of research. But they've given themselves, with EPSRC's backing, the best chance of making some serious progress.

The overall objectives of the IRC in medical image were also included in the report. They are:

… to develop a coherent understanding of the generic IT problems involved in extracting clinically useful information from medical images and signals; pursue a planned programme of IT-centric research aimed at addressing these problems; guide and evaluate the research in the context of a set of realistic clinical exemplars; develop common validation methods and collect large-scale test/validation data sets; and create a coordinated programme of multidisciplinary education and training.

Further details of the theme day can be obtained on: Tel: (UK) 01793 44428; E-mail: j.fleming@epsrc.ac.uk

Table AI. Examples of the inter-relationship between the physical and life sciences