Deep Chemometrics for Nondestructive Photosynthetic Pigments Prediction Using Leaf Reflectance Spectra

Abstract The need for the rapid assessment of the photosynthetic pigment contents in plants has encouraged the development of studies to produce nondestructive quantification methods. This need is driven by the fact that data on the photosynthetic pigment contents can provide a variety of important information that is related to plant conditions. Using deep chemometrics, we developed a novel one-dimensional convolutional neural network (CNN) model to predict the photosynthetic pigment contents in a nondestructive and real-time manner. Intact leaf reflectance spectra from spectroscopic measurements were used as the inputs. The prediction was simultaneously carried out for three main photosynthetic pigments, i.e., chlorophyll, carotenoid and anthocyanin. The experimental results show that the prediction accuracy is very satisfying, with a mean absolute error (MAE) = 0.0122± 0.0004 for training and 0.0321± 0.0022 for validation (data range of 0 – 1).