Non-destructive radio frequency sensor methodology for early nematode detection in walnut roots

Overview

This solution offers a non-destructive, low-cost methodology for early detection and classification of nematode infestation levels in walnut tree roots. The technology addresses a critical pain point in the walnut industry, where late detection of nematode infections leads to significant yield losses and costly remediation. By combining a compact radio frequency tridimensional sensor with machine learning classification algorithms, the approach provides orchard managers with timely, actionable intelligence to guide pest management decisions before infestations become severe.

Technical specifications

Core technology:

• The Walabot, a programmable, pocket-sized radio frequency tridimensional sensor, is positioned proximate to walnut roots to capture radio frequency reflectance signals.
• Reflectance data is processed through machine learning algorithms, including Neural Networks, Support Vector Machines, and Random Forest classifiers, to predict infestation severity.
• Infestation is categorized into four distinct levels based on root-lesion nematode counts.

Validation performance:

• Neural Networks classification model achieved up to 96% accuracy in the MESA lab (Mechatronics, Embedded Systems, and Automation) assessment.
• Field validation across 2019, 2020, and 2021 measured over 100 sampling trees from six different blocks, with each tree measured five times to reduce anomalous results.
• Neural Nets delivered the highest classification accuracy at 92% across multi-year field trials.

Key advantages:

• Non-destructive measurement preserves tree health during assessment.
• Low-cost hardware compared to traditional laboratory nematode extraction and counting methods.
• Portable, pocket-sized form factor enables in-field deployment by orchard personnel.

Technology readiness level

The technology has progressed through multi-year field validation, with data collected across three growing seasons (2019–2021) involving more than 100 walnut trees. The MESA lab at Texas A&M University has demonstrated proof-of-concept performance, with Neural Networks achieving 96% accuracy in controlled assessments and 92% accuracy in field trials. Future work includes expanded field experiments at UC Kearny over a one-year period to collect additional training data, with the goal of further improving model accuracy. The solution is currently at a validation stage, ready for expanded pilot testing and refinement before broader commercial deployment.