文献类型：期刊文献

英文题名：Hyperspectral imaging-based prediction of soluble sugar content in Chinese chestnuts

作者：Yang, Jinhui[1] Gong, Bangchu[1] Jiang, Xibing[1,2]

通信作者：Jiang, XB[1];Jiang, XB[2]

机构：[1]Chinese Acad Forestry, Res Inst Subtrop Forestry, Hangzhou, Peoples R China;[2]Chinese Acad Forestry, State Key Lab Tree Genet & Breeding, Beijing, Peoples R China

年份：2023

卷号：6

外文期刊名：FRONTIERS IN FORESTS AND GLOBAL CHANGE

收录：;Scopus(收录号：2-s2.0-85165109688);WOS:【SCI-EXPANDED(收录号:WOS:001024995300001)】；

基金：Funding This research was funded by the Zhejiang Science and Technology Major Program on Agriculture New Variety Breeding (2021C02070-4).

语种：英文

外文关键词：regression; plant growth; tree production; spectroscopy; Chinese chestnuts; soluble sugar content

摘要：Soluble sugars are critical determinants of fruit quality and play a significant role in human nutrition. Chestnuts, rich in soluble sugars, derive their sweetness from them. However, their content varies with cultivar, location, and environmental conditions. Traditional methods for determining soluble sugar content in chestnuts are time-consuming, laborious, and destructive. Therefore, there is a pressing need for rapid, non-destructive, and straightforward methods for determining soluble sugars in chestnuts to expedite genetic selection. This study aimed to develop a hyperspectral imaging-based prediction model for soluble sugar content in Chinese chestnuts. Firstly, abnormal samples were eliminated using ensemble partial least squares for outlier detection. We then compared the average original and block scale (BS) spectra, with the latter demonstrating significant differences. The BS pretreatment exhibited two small absorption peaks in the 403.7 & SIM; 429.1 nm band and 454.7 & SIM; 500 nm band, less fluctuation in the spectral curves from 503.2 to 687.2 nm, and a substantial increase in spectral absorption between 690.6 and 927.8 nm. Subsequently, we developed a partial least squares (PLS) model using BS pretreatment and regularized elimination (rep) variable selection, which showed better accuracy in predicting chestnut soluble sugar content than other variable selection methods. The model fitting accuracy after the spectra treatment was marginally better than that of the original spectra, with a calibration set correlation coefficient (R2) of 0.59 and root mean square error (RMSE) of 1.02, and a validation set R2 of 0.66 and RMSE of 0.94. The wavelengths at 464.3, 503.2, 539.3, 579, and 711.3 nm were identified as critical for developing the soluble sugar content prediction model. The study demonstrated the potential of Near-Infrared Spectroscopy (NIS) as a rapid and non-destructive method for predicting chestnut soluble sugar content, which could be beneficial for quality control and sorting in the food industry.