From Cows to Crops: Computer Vision for Precision Agriculture
Recent technological advances have allowed a step-change in the way the farmers, agronomists and plant breeders can gather and analyse data, including automated livestock welfare management, precision weed control and measurement of phenotypic traits in plants, allowing greater yields for fewer inputs (feed, agro-chemicals, etc.). Central to many of these systems is the concept of “Computer Vision”, the process of analysing images or videos to automatically obtain meaningful measurements, without the need for manual intervention. This talk will discuss some such systems and how they directly benefit the agricultural and plant science community, using real-world, state-of-the-art examples currently under development.
Ian obtained a PhD from the University of Leeds on the topic of 3D reconstruction in unconstrained pedestrian scenes. He moved to the BRL in 2014 and has since developed significant experience in 3D image capture and analysis, primarily in the Agri-Tech sector. In this vein, he has pioneered the development of several techniques for plant recognition, including in the field of precision agriculture, where he developed a system for automated weed detection in maize crops at real-world speeds and extremely high resolution compared to existing techniques. He has performed investigations into how the cost of this technology can be reduced to allow penetration into new areas of research, with particular application to plant phenotyping.
His work tends to venture outside of the laboratory and into the real-world and he has recently published work on 3D surface reconstruction at extremely long distance and in bright sunlight by exploiting novel sensor technologies in combination with state-of-the-art computer vision techniques.
His most recent project is the development of a system for automated health and welfare management in dairy cattle, again taking advantage of the wealth of rich data that can be garnered from 3D imagery to take measure traits affected by lameness and condition, with the aim of detecting them early and preventing unnecessary suffering, whilst also maximising milk yield.
His key research interests lie in structure-from-motion techniques and 3D morphological analysis for agriculture with application to weed control, fine-grained 3D analysis for phenotyping and plant deficiency monitoring.