Classifying cadmium contaminated leafy vegetables using hyperspectral imaging and machine learning

• DOI: 10.1016/j.heliyon.2022.e12256
• Publication Date: 2022-12-14T17:00:03Z
• Authors: Augusto Souza; Maria Zea Rojas; Yang Yang; Linda Lee; Lori Hoagland
• Keywords: VNIR

Cadmium (Cd) is a toxic element that can accumulate in edible plant tissues and negatively impact human health. Traditional Cd quantification methods are time-consuming, expensive, generate lot of waste, slowing development to reduce uptake. The objective this study was determine whether hyperspectral imaging (HSI) machine learning (ML) be used predict concentrations plants using kale (Brassica oleracea) basil (Ocimum basilicum) as model crops. experiments were conducted an automated phenotyping facility where all environmental conditions except soil concentration kept constant. determined at harvest traditional train the ML models with data collected from sensor. Visible/near infrared (VNIR) images also processed calculate reflectance 473 bands between 400 998 nm. All spectra subject feature selection algorithm ReliefF Principal Component Analysis (PCA) provide input evaluate three classification models: artificial neural network (ANN), ensemble (EL), support vector (SVM). Plants categorized according higher or lower than safety threshold 0.2 mg kg-1 Cd. Wavelengths highest ranks for detection 519 574, 692 732 nm, indicating content likely altered plants' chlorophyll leaf internal structure. able sort into groups, though best F1 score ANN validation subset utilized wavelengths. This demonstrates HSI promising technologies fast precise diagnosis leafy green plants, additional studies needed adapt approach more complex field environments.