Harnessing AI for Real-time Detection and Mitigation of Herbicide-Resistant Pigweeds

Moonshot Award

Hybrid Award (Proposal Development & Innovation Development)

Abstract

Herbicide-resistant waterhemp (Amaranthus tuberculatus) is an emerging threat to soybean production in New York and the northeastern United States, where rapid detection and management are critical to prevent establishment, spread, and long-term economic losses. The goal of this Cornell Moonshot project is to develop AI-enabled scouting tools and decision-support systems that empower growers and crop advisors to detect, monitor, and manage waterhemp infestations at their earliest stages. By integrating advanced sensing technologies, machine learning, and precision agriculture, we aim to enable proactive and site-specific waterhemp management strategies. 

Our research combines hyperspectral sensing and deep learning to improve waterhemp detection and characterization. Hyperspectral analyses showed that glyphosate-resistant and susceptible waterhemp populations become spectrally distinguishable six days after herbicide application, with the strongest discriminatory signals occurring in the red-edge region (700-780 nm). Machine-learning approaches further identified a small subset of informative wavelengths capable of detecting glyphosate resistance, supporting development of cost-effective field sensors. For early detection of waterhemp, a YOLO26n deep-learning model trained on annotated imagery achieved up to 95.1% precision and 93.9% mAP@50 in identifying young waterhemp plants. Building on these results, ongoing work is deploying drones across New York soybean fields to collect high-resolution aerial imagery and develop AI models that detect, map, and track waterhemp-infested hotspots. 

The long-term vision is an operational decision-support platform that enables targeted interventions, reduces unnecessary herbicide applications, slows resistance evolution, and mitigates the establishment of herbicide-resistant waterhemp across northeastern soybean production systems.