Hyperspectral sensing for detecting pesticide residues on crops

Overview

This solution leverages UAV-based hyperspectral imaging and machine learning to enable rapid, non-destructive detection and differentiation of crop foliage based on pesticide residue presence. By analyzing plant spectral signatures from the air, the technology aims to disentangle pesticide-related spectral signals from other environmental and crop management variables such as nitrogen fertilization. The approach supports precision agriculture by providing a scalable method to monitor pesticide coverage and residues across field plots, ultimately helping farmers and agronomists optimize pesticide application, reduce environmental impact, and improve compliance with residue regulations.

Technical specifications

Core capabilities:

• Hyperspectral data acquisition using handheld spectrometers and UAV-based multispectral and hyperspectral cameras for multi-scale sensing from the lab to the field
• Machine learning models trained to classify and detect foliage pesticide residues from spectral signatures
• Ground-truth validation using the QuEChERS extraction method followed by liquid chromatographic-mass spectrometric analysis to confirm residue presence and concentration on leaf samples
• Split-plot experimental design comparing pesticide-treated and untreated plots alongside nitrogen fertilization treatments (0 and 100 kg N ha-1) on winter wheat
• Complementary measurements including above-ground biomass, leaf chlorophyll and nitrogen content, and insect biodiversity counts via insect traps to contextualize spectral findings

Technology readiness level

The underlying hyperspectral imaging and machine learning pipeline has been validated through field plot experiments on winter wheat, demonstrating the ability to capture and analyze plant spectral responses related to nitrogen fertilization and biodiversity. The next phase of validation extends the experiment by introducing pesticide treatment split-plots and collecting leaf residue data via QuEChERS-LC-MS to train and validate residue detection models. The technology is at an early-to-intermediate research stage, with active field validation underway and readiness for collaborative pilot deployment.