Skip to main content
ARS Home » Southeast Area » Tifton, Georgia » Southeast Watershed Research » Research » Research Project #443159

Research Project: High Throughput Phenotyping of Crops Using sUAS-borne Imagery

Location: Southeast Watershed Research

Project Number: 6048-13000-028-015-S
Project Type: Non-Assistance Cooperative Agreement

Start Date: Sep 1, 2022
End Date: Aug 31, 2025

Via cooperative research among FVSU and USDA scientists, this project addresses questions underlying the characterization of phenological characteristics in crops. Phenology and observable phenotypic traits of crops are key features used by plant breeders to assess the genetic make-up & variation and characteristics, including that have potential economic value for developing new breeding lines and cultivars. However, field work involved in painstakingly measuring and recording phenological traits across multiple accessions throughout a growing season is laborious, costly and time consuming. Imaging and subsequent high-throughput analysis of the imagery offers a way to cut some of the time and expense of manual observation. The use of small Unmanned Aircraft Systems (sUAS) to conduct this imaging has become a new area of investigation with the prospect of capturing high resolution imagery, not only in the visible (VIS) spectrum, but at multiple frequencies, by using multi- and hyperspectral sensors. Further information can be derived by capturing high resolution surface characteristics of plant structures as crops mature, using light detection and ranging (LiDAR) technologies.

This project will use sUAS technologies to capture data detailing the hyper- and multi-frequency spectral signatures as well as the physical structure of a panel planting of Sorghum bicolor. Datasets will be collected from hyperspectral and LiDAR sUAS-borne instruments. The goal of this work will be to develop methods for the use of image classification and modeling of the breeding line phenotypes, which can compare or surpass the efficiency of traditional, field methods of phenotyping. To operationalize these methods, classification processes will be conducted using machine learning algorithms on a high-performance desktop computer. Subsequently, once a workflow is established, the process will be repeated and further developed in the ARS high-performance computing (HPC) SCINet clusters.