Dr Julie Bremner (Cefas)
Dr Michal Mackiewicz (School of Computing Sciences, University of East Anglia)
Coastal remote sensing uses aerial images to track the distribution of features such as vegetation, birds, and mammals . Current sensing systems suffer from illumination problems, particularly shadows, which degrade the images recorded. Indeed, the same feature can induce a different image response when viewed in or out of shadow. Human vision overcomes this obstacle by interpreting the properties and proximity of objects in a scene, but remote sensing systems lack the brain’s interpretive power and regularly mis-classify features.
In this project, building on research in the computer vision community, we will develop image processing methods to help us to ‘see into shadows’ and thereby to better classify the coastal features found in images. Our overarching aim is to develop a vision system which – by seeing into shadows – will significantly improve the accuracy of surveys of the coastal environment.
This exciting project will see you working with a team of computing and natural scientists at the cutting edge of remote sensing research. You will characterise the capture device (measuring the camera’s spectral response ) and the statistics of the physical world that is being surveyed, to understand the spectral properties of the coastal environment. You will undertake modelling of the coastal environment in collaboration with Cefas, using a large annotated imagery dataset, with an emphasis on identifying the same image features seen in and out of shadows. You will re-purpose  and  for algorithm development, incorporating our understanding of the camera and the physical world. A particularly novel aspect is that we will incorporate near-infrared into the algorithm formulation. There is evidence  that near-infrared can help distinguish shadowed and non-shadowed areas.
The candidate – who should have a scientific or engineering background/degree – will be involved in all aspects of the project. They will be trained in measurement and calibration at the world-leading UEA Colour Lab and will work directly with Cefas. The developed algorithms will be prototyped in Cefas’ classification framework.