Skip to main content
ARS Home » Midwest Area » West Lafayette, Indiana » Crop Production and Pest Control Research » Research » Publications at this Location » Publication #403748
Research Project: Fungal Host-Pathogen Interactions and Disease Resistance in Cereal Crops

Location: Crop Production and Pest Control Research

Title: The utility of proximal sensing and deep learning in the detection and characterization of tar spot epidemics on corn in the United States

Author
item LEE, DA-YOUNG - POHANG UNIVERSITY OF SCIENCE & TECHNOLOGY
item NA, DONG-NEOP - POHANG UNIVERSITY OF SCIENCE & TECHNOLOGY
item GONGORA, CARLOS - NATIONAL TECHNOLOGICAL INSTITUTE OF MEXICO
item Goodwin, Stephen - Steve
item LEE, JEONG-SOO - POHANG UNIVERSITY OF SCIENCE & TECHNOLOGY
item CRUZ, CHRISTIAN - PURDUE UNIVERSITY

Submitted to: Plant Pathology International Congress
Publication Type: Proceedings
Publication Acceptance Date: 5/10/2023
Publication Date: 8/20/2023
Citation: Lee, D., Na, D., Gongora, C., Goodwin, S.B., Lee, J., Cruz, C.D. 2023. The utility of proximal sensing and deep learning in the detection and characterization of tar spot epidemics on corn in the United States. Plant Pathology International Congress. ABSTRACT.

Interpretive Summary: N/A

Technical Abstract: In recent years, the widespread incorporation of image sensors, whether through proximal or unmanned aerial remote sensing, has proven their utility as an alternative approach to conventional, human-vision-based disease estimation. Nevertheless, obtaining objective, accurate, and high-throughput measurements of signs and symptoms of a plant disease from its early to late stages are the most important priority in sensor-based phenotyping. Since its first identification in 2015 in the U.S., tar spot of corn caused by Phyllachora maydis has rapidly spread from Illinois and Indiana through the corn belt and south to Florida. The detrimental impact on yield and the polycyclic nature of tar spot epidemics have made this disease one of the most significant emerging diseases of corn in the United States. In my talk, I will share our work towards developing a pipeline that consists of the previously developed Stromata Contour Detection Algorithm (SCDA v1) and the generation of a Convolutional Neural Network (CNN). Our approach allows high-throughput and automated detection of tar spot stromata in Red-Green-Blue (RGB) images of corn leaves collected at multiple experimental sites in Indiana in 2021 (onset to later stages of tar spot development). Our work will serve as a foundation for building an accurate and reliable standardized approach that can be utilized nationally and internationally for tar spot research, disease management, surveillance, and epidemiological modeling.