Year Identification of Seeds in Peony (Paeonia suffruticosa Andr.) Using Hyperspectral Imaging

Abstract

Seed storage year is one of the important indicators for evaluating the quality of peony seeds. It is of great significance for the development of the peony industry to carry out rapid and non-destructive year identification of peony seeds to provide a basis for the screening of aged seeds during seed breeding and processing. This study explores the feasibility of using hyperspectral imaging technology combined with machine learning methods to identify the two states of peony seeds (shelled and non-shelled) and then determines the most suitable state for the year identification of peony seeds. The two states of peony seeds (shelled and non-shelled) in 2017, 2018, and 2019 are employed as the research objects. Hyperspectral imaging data of two kinds of peony seeds in the spectral range of 935-1720 nm are collected. The machine learning methods based on the two states of peony seeds (shelled and non-shelled), including partial least squares (PLS-DA), Support Vector Machine (SVM), and Convolutional Neural Network (CNN) classification models, are established and compared. It is found that the optimal year identification models of peony seeds (shelled and non-shelled) based on hyperspectral imaging technology have better recognition effects and the recognition accuracy is more than 99.5%. Moreover, the recognition accuracy of the year identification PLS-DA model established by non-shelled peony seeds is 99.96%, which is better than that of shelled peony seeds at 99.64%. This indicates that year identification of peony seeds based on hyperspectral imaging technology is feasible and efficient and that non-shelled peony seeds are more suitable for the year identification of peony seeds. The results can provide a theoretical and methodological justification for the screening of high-quality peony seeds.