Deep Learning for the Identification and Quantification of Wheat Disease

Long, Megan (2022) Deep Learning for the Identification and Quantification of Wheat Disease. Doctoral thesis, University of East Anglia.

Preview

PDF Download (22MB) | Preview

Abstract

Wheat is hugely important across the globe, providing food and nutrients for millions of people and livestock. Like all crops, it struggles with pressure from multiple diseases, which need to be controlled to preserve both yield and quality. Breeding new varieties with resistance to important diseases is a long process that takes many years and requires trained pathologists to manually score thousands of plots for disease levels. Automating the disease scoring process would free up time for pathologists to work on other important tasks. It also has the potential to improve the accuracy of scoring through multiple applications and eliminating human error.

Here we present a dataset of wheat images taken in real growth situations, including field conditions, with five categories: healthy plants and four foliar diseases, yellow rust, brown rust, powdery mildew and Septoria leaf blotch. This dataset was used to train deep learning models to identify and classify the diseases. We collect a quantification dataset of yellow rust images and performed experiments with different score categories to train various models. Finally, we carry out experiments with simulated data for determining the viability of deep learning models for disease quantification.

In this thesis we find that deep learning models are capable of classifying complex field images, with accuracies of over 97%. We identify limitations in the data collection for quantification of wheat diseases in the field and provide a method for determining ideal dataset size. These results show the viability of deep learning models for quantifying disease and determine some of the challenges which need to be overcome to develop an automated method for use in the field.

Item Type:	Thesis (Doctoral)
Faculty \ School:	Faculty of Science > School of Biological Sciences
Depositing User:	Chris White
Date Deposited:	05 Sep 2023 10:24
Last Modified:	30 Sep 2023 01:38
URI:	https://ueaeprints.uea.ac.uk/id/eprint/92990
DOI:

Downloads

Actions (login required)

View Item