INVITED KEYNOTE
Applications of hyperspectral imaging from planetary science through to healthcare
Prof Mark R Sims, Space Research Centre, Leicester University.

SPEAKER BIO

Leader of the SRC's planetary research and technology programme and associated spin-off (knowledge exchange) activities.
University of Leicester Science Faculty Lead for Diagnostics Development Unit (www.le.ac.uk/ddu ) now commissioned and collecting data at Leicester Royal Infirmary which uses state of the art instruments to attempt non-invasive diagnosis of disease. Includes use of multi-spectral/hyperspectral imaging derived from SRC’s planetary programme.

Chairman UK’s STFC Aurora Advisory Committee and delegate to the ESA HME Programme board 2004 to December 2009. Member of the Technology Strategy Board’s National Space Technology Steering Group. Member of the UK Space Agency’s Space Technology Advisory Committee.

Research interests include space technology, instrumentation, image processing, optical system design and data processing. Particularly interested in using instrumentation and technology in interdisciplinary research to achieve new novel and perhaps unconventional solutions to research challenges and in particular currently the challenge of detecting Life in extreme environments and on other planets such as Mars and using non-invasive or non-destructive techniques for detection of disease and forensic evidence.

Interested in science education and in particular science education within primary schools.

Principle Investigator for Life Marker Chip Instrument proposed for ExoMars and other Mars Missions.

Beagle 2 Mission Manager responsible for instrumentation and flight operations implementation and test on the Beagle 2 Mars Lander probe launched on ESA’s Mars Express Mission in 2003. This was Europe’s first mission to the planet and the UK’s first planetary lander. Led Flight Operations and Beagle 2 team’s investigation into failure of mission.

Spatial/spectral or colour classification a quantitative analysis
Prof Steve Marshall and Tim Kelman, HSI Centre, University of Strathclyde.

ABSTRACT

This presentation quantifies the benefits of using a Hyperspectral camera over a standard colour RGB camera, with and without spatial features for several industrial applications.

Off the shelf development kits are readily available which allow vision systems with standard RGB cameras to be implemented quickly.  Such vision systems are able to carry out a variety of inspection tasks using spatial and colour features.  Currently most of these systems do not normally incorporate Hyperspectral sensors.

A question posed by most business managers is, ‘what  improvement  will I get by investing in HSI technology?’.

This presentation considers the sorting of a range of products, drawn from the food sector, such as different types of peanuts and rice. We compare the recognition rates using features derived from spatial, RGB colour and hyperspectral  data plus combinations thereof.

The key comparison comes between (Spatial + Hyperspectral) v (Spatial + RGB colour). This is the really important comparison driving the industry business decision.

It informs the industry what more they will get if they invest in an HSI camera as opposed to just using a standard colour camera.

This work gives a quantitative analysis which shows the improvement in classification accuracy achieved by using HSI for different industrial problems.

SPEAKER BIO

Tim Kelman received an MEng with Distinction in Electronic and Electrical Engineering from University of Strathclyde in 2010.  He is currently studying for a PhD in the Hyperspectral Imaging Centre at University of Strathclyde and is sponsored by Strathclyde University and Gilden Photonics. He is working in the area of novel classifiers for HSI applications. Last year Tim spent three months at Beijing Jiatong University in PR China.

Professor Steve Marshall is Head of the Signal and Image Processing group in the Dept of Electronic and Electrical Engineering Department  at University of Strathclyde. He received a 1st class honours degree in Electrical and Electronic Engineering from University of Nottingham and a PhD in Image Processing from University of Strathclyde.  He has published over 200 conference and Journal papers. His interests are signal and image processing and in recent years he has applied these skills to hyperspectral data and established the Hyperspectral Imaging Centre at University of Strathclyde in collaboration with Gilden Photonics and other industrial partners. The centre carries out projects of applied research on industrial applications of HSI specialising mainly on the data processing aspects.

The high throughput virtual slit: A flux multiplier for high resolution hyperspectal imaging
Dr Arsen Hajian, Chief Technology Officer, Tornado Spectral Systems, Canada.

ABSTRACT

Tornado Spectral Systems has commercialized a new kind of hyperspectral imaging platform technology that relaxes the tight coupling between the spectral and spatial resolution imposed by the slit in a conventional hyperspectral imager. Known as the High Throughput Virtual Slit (HTVS), this component anamorphically reformats the input source image of a broadband light spectrometer to massively increase the photon flux on the focal plane at a given spectral resolution.

We consider this technology in the context of high spectral resolution applications such as Raman imaging. In this case, conventional hyperspectral systems require a narrow slit to provide high spectral resolution, but resulting system throughput is low. With an HTVS-equipped hyperspectral imager, we show that the signal incident on the detector can be increased by a large factor while maintaining high spectral resolution. The HTVS technology is based on reflective optical elements, does not introduce any additional limitations to the performance of the hyperspectral imager, and can be incorporated into existing designs. We will discuss how the introduction of this fundamental technology changes the metrics for optimal hyperspectral imaging design.

ABSTRACT

Techniques that are ables to record spectral or hyperspectral video sequences are notable by their scarcity. Fundamental reasons for this include the problems of recording a three-dimensional spectral data cube with two-dimensional detector arrays; optical throughput and the space-bandwidth product of detector arrays. This talk will address modern approaches to addressing these problems including research at Glasgow University into video-rate and random-access spare spectral imaging.

SPEAKER BIO

Prof Andy Harvey obtained his PhD from St Andrews University in 1990 where he subsequently conducted post-doctoral research in fusion plasma diagnostics and optical instrumentation. He moved to DERA (now QinetiQ) Malvern in 1995 to work on various imaging techniques including aperture synthesis, thermal imaging, adaptive optics and fibre optics. Following  two at Cranfield University he moved in 2001 to Heriot Watt University where he established the Imaging Concepts Group conducting research in retinal imaging, wavefront coding, hyperspectral imaging, and aperture synthesis. He has published over 100 journal and conference papers and six patents. He is a recent chair of both the Optical Group and the Optics and Photonics Division of the Institute of Physics. In January 2012 he took up a new Chair in Optics in the Department of Physics at Glasgow University where he continues his research in advanced imaging technique

Mid wave IR/online applications
Timo Hyvärinen, SPECIM Oy, Finland

Improving the usefulness of colour in human and machine vision through
hyperspectral imaging
Prof Anya Hurlbert, Institute of Neuroscience, University of Newcastle.

ABSTRACT

1. To identify new spectral signals useful for specific visual tasks, such as object recognition. Example: to identify novel spectral signatures of natural objects such as human skin.
2. To simulate natural illumination changes and thereby create novel images for use as visual stimuli in evaluation of human performance.
3. To provide more useful colour image databases for computational modelling and testing of colour constancy and colour correction algorithms.
   
   

SPEAKER BIO

Professor Anya Hurlbert (ACH) is founding Director of the Institute Neuroscience and Professor of Visual Neuroscience in the Faculty of Medical Sciences at Newcastle University. Trained as a physicist (BA, Physics, Princeton University), physiologist (MA, Cambridge University), neuroscientist (PhD, Brain and Cognitive Sciences, MIT), and physician (MD, Harvard Medical School), she combines behavioural, computational, and imaging approaches to understanding human vision.

Her research has been funded by the SERC/EPSRC, MRC, the Wellcome Trust, Unilever, Nissan, the EU, and others. Main areas of research are in : developing and testing computational models as explanations of colour perception, particularly focussing on physical reflection effects in complex, natural scenes; understanding variations in colour perception and preference across ages and populations; understanding the neural mechanisms of human colour perception. 

Her publications include original articles in Nature, Science, and Current Biology, and Proceedings of the National Academy of Science and numerous book chapters and invited papers. She is past Chairman of the Colour Group (GB), on the editorial board of Current Biology, member of the Visiting Committee for MIT’s Department of Brain and Cognitive Sciences, and Scientist Trustee of the National Gallery (London). She is active in public engagement and media activities, including many events on the use and meaning of colour (e.g. ‘Cracking the Colour Code’, Gedeon Programmes, 2008; ‘Do you see what I see?’, BBC Horizon, 2011).

Light paths in the retina and its implications for hyperspectral imaging of oxygen saturation
Prof Steve Morgan, Nottingham University.

ABSTRACT

Accurate oxygen saturation imaging of retinal blood vessels via hyperspectral or spectral imaging is difficult to achieve without knowledge of the path lengths that the light takes within the vessel. A Monte Carlo simulation of light propagation through the retina has been developed to understand the path length distributions within the retinal vessel and fundus. For full-field illumination, the path length distribution within the vessel comprises directly backscattered light and light that has passed once or twice through the vessel. The origins of these light-path-length distributions can be better understood by investigating different combinations of single-point illumination and detection positions. The most significant observation is that illumination from the side of the vessel and detection directly above the vessel extracts only the light that has taken a single pass through the vessel. This path length distribution is tightly constrained around the diameter of the vessel and can potentially provide enhancements for oxygen saturation imaging. The method could be practically implemented using an offset-pinhole confocal imaging system or structured light illumination.

Multispectral remote sensing of blood oxygenation
Dr Peter Yuen, Cranfield University.

ABSTRACT

Hyperspectral imaging (HSI) has been one of the emerging techniques for assessing tissue oxygen saturation (StO2) in-vitro remotely. Current methods using monochromatic LED light sources or optical fibre probe could only assess local regions of body tissue, while HSI has the capability to probe large areas of StO2 distribution across the whole body. Previous HSI work in the field has commonly adopted a few, or, a range of specific wavelengths, within the visible region (VIS) of 550-600nm or in the near infra-red (NIR) of 600-1050nm wavelengths using various forms of Beer-Lambert (BL) formulations for the StO2 assessment. However, none of the reported work ever attempted to elucidate why these wavelength were chosen, nor is there any information whatsoever to explain what are the effects for selecting the wavelengths other than their chosen ones. Furthermore, various BL models have been employed and it is not clear what are the advantages and drawbacks for using these models particularly in the remote sensing of StO2. This presentation gives an overview of why some wavelengths are more suitable than others for the StO2 assessment, and to highlight other factors such as specular reflections from the skin which may mislead the interpretation.

Peter Yuen is fellows of Institute of Physics (FInstP) and Institute of Mathematics and its applications (FIMA) since 2001 and 2002 respectively. Peter obtained his Ph.D in semiconductor physics from Imperial College, University of London in 1987 and subsequently he worked in this field till 1997. Then he joined the Asahi Chemical Industry Co. in Japan for two years to develop narrow gap semiconductor devices for magnetic sensing applications. In 1999 he joined the UK defence agency DERA as a principal scientist in the area of underwater defence research. In 2003-2007 he was a senior principle scientist with the BAE Systems UK for the development of hyperspectral remote sensing technology. In May 2007 Peter joined Cranfield University as a senior lecturer to lead a group on Electro-optics, remote sensing & computer vision research. Peter has about 75 journal and conference papers in the fields of semiconductor physics, computer vision and defence science.

Mapping food composition by NIR hyperspectral imaging
Dr Martin Whitworth, Principal Scientist - Cereal Science and Technology, Campden BRI.

ABSTRACT

Near infra-red spectroscopy is widely used in the food industry for rapid on-line and laboratory analysis of composition.  Established applications include fat content in meat, and moisture and protein analysis in grain and flour.  Hyperspectral NIR imaging provides the industry with new opportunities to study the distribution of such components in foods, with applications in product development and quality control.  Areas of interest include grading of meat, study of moisture migration between food components during storage, and fat uptake in frying processes.

Campden BRI have used a laboratory system with an HgCdTe camera to demonstrate and evaluate several applications of the method for food industry clients.  The system has a wavelength range of 1000-2500nm and a range of magnifications suitable for sample widths of 10 to 300mm.  Samples are presented on a translation stage and imaged with a push-broom configuration, enabling potential for on-line use to be demonstrated.  Applications will be presented including measurement of moisture distribution in bread, during baking of digestive biscuits, and between the pastry and filling of custard tarts, fat distribution in chips, beef quality, fish freshness, and detection of added gluten in flour.

Martin Whitworth is a Principal Scientist for Cereal Science and Technology in Campden BRI’s Cereals and Milling Department.  He has worked at Campden BRI and its predecessors since 1992.  He has an MA and PhD in physics from Cambridge University, and is a member of the Institute of Physics and a Chartered Physicist.  He won the silver medal in 1994 from the National Association of British and Irish Millers as the highest placed student in their flour milling correspondence course, and won the Heinz travelling scholarship in 1995 for his work on fracture mechanics and image analysis.

Martin’s research interests include the use of imaging techniques and digital image analysis for food analysis, the bubble structure of dough and baked products, analysis of food colour and structure and the mechanical properties and size of cereal grains.  Applications include the use of X-ray computed tomography to study bread and cake structure during baking, calibrated colour imaging for food specification and quality control, and hyperspectral NIR imaging to measure the distribution of components such as moisture and fat in food products.  Martin has developed several practical applications for the food industry including an on-line system for measurement of bran in flour which formed the basis for the commercial Branscan and Fluoroscan instruments, and the C-Cell instrument for measurement of bread quality.

Martin has presented work at many food industry conferences and Campden BRI courses.  He has received awards from the AOCS for an Outstanding Presentation at their 2001 conference and from the Rank Prize Funds Opto-Electronics Committee for the best contributed paper at a mini-symposium in 2004.

SPEAKER BIO

Paul Murray recently completed a PhD at the University of Strathclyde in Glasgow where his work focused on extending the standard Hit-or-Miss Transform for application to noisy data. He is currently working as a Research Associate at the University of Strathclyde. His research interests lie in using hyperspectral imaging and advanced signal and image processing techniques to study pharmaceuticals, biological processes, and a number of problems encountered in the food and drink industry.

Recent industrial applications of novel NIR hyperspectral inspection technique that overcomes x-ray blind spots
Dr Geoff Diamond, CTO, Inspection Technologies

ABSTRACT

Results from a recently commercialised Near Infra Red (NIR)) defect-detection system for the inline inspection of foods is presented. The system identifies the presence of materials in foodstuffs that are normally transparent to X-rays and hence undetectable via conventional inspection processes. The system operates at low voltages, is completely non-ionising and can easily be retro-fitted into existing production lines and can operate at belt speeds of upto 60 metres per minute.

Typical radiolucent foreign objects that are common problems in the food industry and which can readily be detected by this new NIR systems are fruit pits and stones, plastics, paper, wood, and insects. Results are also presented from the meat processing sector where the inability of x-ray systems to detect cartilaginous tissues and non-calcified bones in chicken breast has been a longstanding industry problem.

Please register here.