Diagnosing fertilizer levels in spinach and detecting pesticides in cauliflower using an electronic nose augmented with machine learning

Abstract The increasing use of pesticides and fertilizers in agricultural practices poses substantial consumer risks. This study explores the utilization of an E-Nose (electronic nose) system augmented with machine learning (ML) to identify fertilizers and pesticides in plants. E-Nose technology effectively analyzes the volatile organic compounds (VOCs) emitted by plants at different conditions. The principle behind E-Nose technology involves using a gas sensor array to record and monitor VOC emissions from plant samples. These emissions are then analyzed using ML models to identify patterns related to pesticide and fertilizer treatments. One significant advantage of this technology is its capability to detect chemical residues without direct contact with the plants, making it a safer and more efficient alternative to traditional chemical analysis. The scope of application of E-nose technology extends to various agricultural monitoring requirements, particularly in identifying the fertilizer and pesticide applications in plants. In this study, the spinach plants were classified into three categories-organic, lightly fertilized, and heavily fertilized. Similarly, the cauliflower plants were exposed to different concentrations of a pesticide mixture containing 40% profenofos and 4% cypermethrin. VOC readings were taken after specific growth periods to capture the effects of these treatments. The generated data was classified using ML models, SVM (Support Vector Machine) and random forest, achieving accuracies of 90.9% for pesticide classification and 99.5% in distinguishing plants treated with fertilizers of different concentrations. The E-Nose technology offers a rapid and cost-effective solution for real-time agricultural monitoring.