Machine Learning Models for Classifying Tomato Plants Exposed to Metal Nanoparticles Through Reflectance Spectroscopy

Nanotechnology has recently found applications in agriculture to optimize crop production, but the contact with plants may also occur as a result of the accidental introduction of nanoparticles into the environment. In this sense, the development of reliable and nondestructive methods capable of identifying and distinguishing plants exposed to nanoparticles is becoming of interest. In this study, machine learning models are used to classify plants irrigated with water containing gold and silver nanoparticles. k-nearest neighbors, random forest, and support vector machine models were trained and tested with the reflection spectra of leaves in the 400–800 nm range. The support vector machine model achieved the best performance, reaching an F1-score above 0.99, correctly classifying 483 of 486 samples in the test set. The findings demonstrated the effectiveness of machine learning, combined with leaf reflectance spectroscopy, in identifying plants exposed to gold and silver nanoparticles. The method may assist the study of the nanoparticles effects in plants development representing a powerful tool in agriculture.