Nir-based aerial phenotyping for early detection of plant pest and pathogen infestations

Overview

This solution uses near-infrared (NIR) reflectance sensors mounted on small unmanned aerial vehicles (sUAVs) to detect pre-symptomatic plant stress caused by insect pests and pathogens. Infested plants exhibit altered biochemical profiles before visual symptoms appear, and these changes produce distinct NIR spectral fingerprints. Machine learning models trained on these fingerprints can classify plants as healthy, mock-inoculated, or infested, enabling growers to identify emerging infestations early, apply targeted treatments, and detect breakdown of host resistance in bred or genetically engineered crops.

Technical specifications

• NIR reflectance sensing captures biochemical signatures of plant foliage that change in response to pest or pathogen attack before visible symptoms develop
• sUAV-mounted sensors are positioned directly in contact with plant canopy for high-resolution field-level data collection
• Machine learning classification distinguishes naïve, mock-inoculated, and infested plants based on spectral profile differences
• Validated accuracy: 86.1% overall testing accuracy for binary classification of mock-inoculated versus inoculated rice plants (N = 72); 73.3% accuracy for three-class classification including healthy controls (N = 105)
• Multi-crop applicability: demonstrated on rice sheath blight (Rhizoctonia solani); planned validation on corn with fall armyworm (including Bt versus refuge comparisons) and wheat rusts (leaf, stripe, stem)
• Temporal profiling capability to track spectral signature changes across varieties with different physical characteristics before and after symptom development

Technology readiness level

The technology has been validated in controlled settings with rice inoculated with Rhizoctonia solani, demonstrating that pre-symptomatic detection via NIR reflectance and machine learning is feasible. The research team has also confirmed that NIR sensors can be effectively mounted on sUAVs for direct foliage contact in field conditions. The next phase of validation during the 2021 field season will test the phenotyping system in open-field settings with corn/armyworm and wheat rust in Ohio, comparing spectral profiles across Bt and refuge corn varieties with and without infestation, and quantifying temporal changes in rust resistance across multiple wheat varieties.