Location: Cereal Disease Lab
Project Number: 5062-21220-023-32-S
Project Type: Non-Assistance Cooperative Agreement
Start Date: Jul 1, 2020
End Date: Jun 30, 2021
The goal of this project is to develop high-throughput phenotyping protocols for rust disease resistance in small grains using remote sensing and big data approaches.
The cooperator will focus on plant stress detection using airborne platforms, and developing Artificial Intelligence algorithms and applying them in various research projects for row and tree crop management. Together, we plan to build on combined expertise on developing high-throughput phenotyping protocols for fast selection of new varieties for rust resistance. New cultivars will be planted in the nurseries located in St. Paul, Minnesota and Rosemount, Minnesota, and trial plots will be planted using complete randomized block design with replications. Our approach is to conduct cereal rust disease detection by color and hyperspectral imaging and various big data techniques, especially by applying deep learning algorithms such as deep convolutional neural network and fast object detection algorithms. Ground truthing data including rust severity by two scoring schemes will be recorded. We plan to improve these algorithms to better work for rust disease detection by collecting large image datasets from multiple platforms, flying altitudes and angles, so that they can be evaluated and widely adopted for various case scenarios that researchers are facing in their current breeding projects. As we analyze large hyperspectral image sets in the projects, we will utilize the Cooperator’s supercomputing facilities for image processing. In addition, a pheno-cart capable of real-time scoring of stem rust will be developed by integrating the sensors, detection algorithms and on-board microprocessors. To enable large-scale experiments, drone path planning algorithms will be carried out to maximize the field coverage from single drone flights.