Modern precision agriculture is increasingly using spatially ordered, high resolution, time series imagery to inform crop management and yield prediction models. This project will build and test an image analysis software module that extracts pertinent features of flowering and grain formation from real-world, GPS-tagged, time series field imagery.
Department: Agri-Environment
Supervised by: Donal O'Sullivan
Agricultural and informatics researchers from SAPD and SSE combined forces in 2015 to create a novel semi-automated instrument driven by bespoke software that acquires and processes overhead RGB image and RTK GPS data acquired in crop fields and stores the processed image, date/time and location data on a remote database server. The instrument has allowed a significant breakthrough in crop genetics and breeding research whose success depends on gathering robust, detailed performance data on thousands of varieties in parallel. This project aims to expand the scope of image processing incorporated in this automated data acquisition pipeline from analysis of its initial target – the wheat leaf canopy - to include also the reproductive structures – the ears. Two critical yield-determining traits related to the ears will be targeted – 1. The ear density (number per unit land area) and 2. Flowering date. The second trait will be approached in two parts. A wheat ear can be determined to have commenced self-fertilization when the yellow pollen-bearing anthers first ‘extrude’ from the green florets, so a generic ability to quantify anther extrusion will be the first step. In order to infer the mid-flowering date, it will be necessary to interpolate from multiple time series images of the same plot. The student will have the opportunity to discuss how the critical traits are recognised and evaluated in the field with the crop research team, receive training and guidance in image analysis and will be supported by a multi-disciplinary research team in the testing and validation of the software solutions he/she comes up with.
The student will first gain an understanding of the existing field imaging/mapping platform through hands-on use and the structure and end use of the acquired field phenotyping data and study prior examples of how 2D RGB image analysis is employed in crop science. The next step is to develop one or more image processing algorithm that extract targeted biological features with minimal false positive and false negative interference. The performance of these algorithms will be benchmarked against existing ground truth data. If time permits, refinements looking at how analysis of time series images of the same physical field location can be employed e.g. to infer date of maximum anther extrusion will be investigated.
Generic algorithm development and coding skills (essential) Knowledge of image analysis algorithms (desirable) Java coding skills (desirable) Interest in data analysis applications in the world of agriculture/food security (desirable)
The student will develop the skills to critically analyse complex real-world user requirements and to work through the software design circle with requirement analysis, design, implementation, and testing. He/she will develop further proficiency in Java coding and a detailed technical knowledge of image analysis algorithms due to synergism of Dr Wei in SSE with the world-leading expertise in SAPD in image analysis pipelines for distinguishing plant features in complex crop canopies (this research is currently subject to a patent application pending possible commercialisation). He/she will also develop team working and time management skills necessary to succeed in liaising with a multi-disciplinary team across two Schools and to achieve tested software products that meet specification within a time-bounded project. If successful, the software developed will likely form a discrete component of a high impact publication reporting the genetic conclusions of the large-scale field experiment, which it is expected that the student will co-author.
The place of work will be on Whiteknights campus in the Agriculture Building, apart from two supervised field trips to the Universities Sonning Farm to familiarize the student with the imaging equipment in use in the field and training sessions based in SSE
Flexible
Wednesday 15 June 2016 - Sunday 21 August 2016
Students should apply by sending a covering letter outlining their preferred destination after graduation and those programming (essential), image analysis (desirable) and any other skills or interests they would highlight as relevant to the project. Please contact either Prof. Donal O'Sullivan (SAPD), d.m.osullivan@reading.ac.uk or Dr. Hong Wei (SSE) h.wei@reading.ac.uk for informal discussion.