The emerging technology of hyperspectral imaging is currently having a major effect on the field of digital imaging. Whereas standard digital camera technology captures a single brightness or colour value at each pixel, hyperspectral cameras capture a complete region of the electromagnetic spectrum. This means that they are able to detect changes in temperature, moisture and fat content and some aspects of the chemical composition of objects in the field of view.
The research and commercial potential of this technology is enormous, offering a wide range of applications, from measuring the degree of absorption of creams on the skin and assessment of burns, to modelling the progress of drug migration through the human body. Other uses include industrial inspection, defence applications as well as detecting counterfeit manufactured goods such as spirits, perfumes and tablets. Hyperspectral imaging is providing new solutions to industrial problems and avoiding expensive, time consuming destructive analysis techniques.
Until quite recently the use of hyperspectral imaging was limited to military and remote sensing applications due largely to the size and cost of the equipment required. These reductions have led to a rapid growth in the use of HSI technology in many applications in lab based environments including inspection in manufacturing, food and drink, agritech, forensics and pharmaceuticals.
HSI requires a hyperspectral camera which typically operates in either the visible, near infra-red or short wave infra-red region. A hyperspectral image therefore takes the form of a cube of data with two spatial dimensions and one wavelength. By their nature HSI data sets are very large occupying hundreds of Gigabytes, though the information content of the data sets can be concentrated in localized areas with the rest of the dataset being sparse. HSI therefore presents new signal processing challenges in terms of data reduction, feature extraction, classification and regression algorithms for interpreting the data. The emergence of deep learning approaches is also beginning to impact on this field with more automated decision making.
The seminar will provide an introduction and overview of hyperspectral imaging followed by some examples of new signal processing advances in this area as well as a series of industrial applications from a range of market sectors.

Speaker(s): Prof. Stephen Marshall,

Location:
College of Engineering and IT Lecture Theatre 102
University of Dubai
Dubai, United Arab Emirates