ASABE Technical Library - Abstract

Click on “Download PDF” for the PDF version or on the title for the HTML version.
If you are not an ASABE member or if your employer has not arranged for access to the full-text, view access and subscription options

An Intelligent Diagnosis and Advisory System for Wheat Stripe Rust

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org

Citation:  2026 ASABE Annual International Meeting  2600103.(doi:10.13031/aim.202600103)
Authors:   Yi-Teng Wu, Qi-Yuan Zhang, Hao-Jie Chen, Zhan-Hong Ma, Ce Yang, Wen-Hao Su
Keywords:   Edge computing; Instance segmentation; Large language model; Lightweight network; Severity grading; Wheat stripe rust.

Abstract. Wheat stripe rust severity grading is essential for timely field management, yet existing monitoring systems often struggle to balance accuracy, edge-side efficiency, and practical decision support. This study presents an integrated real-time diagnosis system combining lightweight visual perception, embedded deployment, and LLM-based reasoning. A YOLO-SSC instance segmentation model was developed to extract wheat leaf regions under complex field conditions, achieving 98.85% mAP50 with 2.37 M parameters and 80.27 FPS. Based on segmented leaf and mask inputs, WheatSeverityNet was designed for six-level disease severity grading, reaching 83.38% accuracy with only 0.124 M parameters and a 0.54 MB model size. The system was deployed on a Raspberry Pi 5 handheld terminal and achieved 4.59 FPS. An LLM-based module further transformed visual results and agronomic context into field management recommendations.

(Download PDF)    (Export to EndNotes)