Predicting internal parameters of kiwifruit at different storage periods based on hyperspectral imaging technology

A new methodology based on hyperspectral imaging (HSI) technology was proposed to predict the soluble solid content (SSC) and the hardness of kiwifruit at different storage periods. Firstly, 240 samples were divided into three groups of 80 samples each, and stored in a freezer for 1 day, 7 days and 14 days. Then, the hyperspectral images of the samples were collected, and the spectrum data in the wavelength range of 400 nm ~ 1000 nm was selected as the full-spectrum data. Secondly, the full-spectrum data was preprocessed by six methods and the Exponential Smoothing (ES) was selected by comparison of the prediction results. Thirdly, the ES-preprocessed spectrum data was extracted by eight methods, and a partial least squares regression (PLSR) model was established to predict the hardness and the SSC of the samples at three different storage periods. The experimental results show that CARS-Boss-PLSR as a combined method, that is, the features extracted by CARS are further extracted by Boss and then they are predicted by the PLSR model, can obtain the optimal results for predicting kiwifruit hardness, with the determination coefficients \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R_{p}^{2}$$\end{document} of 0.88, 0.89 and 0.89, and with the residual predictive deviation (RPD) of 2.92, 2.99 and 3.05. CARS-Boss-PLSR also has the optimal prediction results for predicting kiwifruit SSC, with the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R_{p}^{2}$$\end{document} of 0.88, 0.87 and 0.88, and with the RPD of 2.72, 3.11 and 3.00. The prediction results of CARS-Boss-PLSR are tested for significance by ANOVA, and the results shows that CARS-Boss-PLSR as a combination method is reliable and stable for predicting the internal parameters of kiwifruit at different storage periods.