A Low-Cost Proximate Sensing Method for Early Detection of Nematodes in Walnut Using Machine Learning Algorithms

This chapter presents an innovative low-cost proximate sensing method designed for the early detection of nematodes in walnut trees, leveraging machine learning algorithms. The chapter commences with an introduction, highlighting the significance of early detection in managing nematode infestations and introducing the approach’s cost-effective nature. The Materials and Methods section outlines the study area, emphasizing the use of reflectance measurements with a radio frequency sensor for data collection. The ground truth data collection and processing methodologies are detailed, providing a transparent overview of the experimental setup. The chapter introduces classification algorithms from scikit-learn, along with the application of deep neural networks (DNNs) using TensorFlow. The Results and discussion section follows, including data visualizations for Project 45 in 2019, the performance of classifiers in both Project 45 instances in 2019 and 2020. The findings offer insights into the effectiveness of the low-cost proximate sensing method for early nematode detection, providing a foundation for further exploration and refinement. The chapter concludes with a comprehensive summary, encapsulating key findings and implications of the low-cost sensing approach.