Seeing Spectral Signatures

点击翻译按钮获取中文摘要

There are many application in agri-food production where it would be very useful to measure the spectrum of light reflected from a scene rather than measuring a camera RGB. Indeed spectral signatures are useful for identifying plant species, monitoring growth and to determine the presence of disease, among other tasks . Unfortunately, spectral measurement devices are expensive, cumbersome to use, and often unsuitable for deployment in the field. In contrast conventional RGB cameras are cheap and reliable and are easy to use outside the lab.The core aim of this project is to develop algorithms for mapping the RGB signal recorded by a conventional camera to corresponding spectral signatures that are useful for solving classification problems including ripe vs non-ripe and the diagnosis of plant diseases. The research will be based on calibration, measurement and inference . That is, we will understand the physical characteristics of our device, the spectra in the world we would like to estimate and then we will develop algorithms for mapping the RGB camera measurements to spectra. Our work will also consider mapping RGB+Near Infra Read camera systems and develop algorithms for predicting spectral signature that extend to the NIR part of the spectrum. The research will be developed in close collaboration Antobot The project will dually focus on the development of robust algorithms and the engineering of practical systems that are deployed in the field. At all stages the project will take place in collaboration with Antobot. The combintion of spectral measurement, computer vision and engineering in the context of Antobot's robotic system makes for a truly exciting collaboration that will equip the student with a unique skillset bridging traditional research led computer /color science and engineering and provide springboard for a future inter-disciplinary career in the Agriforwards area. Throughout the project the student will participate in UEA's doctoral College - including its training programme - and the cohort development activities of the Agriforwards CDT.The UEA+Antobot team will meet regularly to review progress and plan for the near and medium term (thus providing an opportunity to identify and then manage risk). The UEA team have a long track record in helping industrial partners to better use their vision systems. Not only does this track record serve to mitigate risk but we will be able to call on our network to help advise on unforeseen problems.