Abstract：Pear is a very common fruit in life and one of the three major fruits in China. The sugar content, acidity (pH) and hardness of the pear are important indexes to evaluate the quality of the pear. Near infrared spectroscopy (NIR) is widely used to detect the quality of fruits because of its fast, non-destructive and high efficiency advantages. The hand-held near-infrared spectrometer can be applied to the on-site nondestructive testing of pear quality. Different pear sizes will have a certain impact on the spectrum and modeling of pears. The near-infrared spectra of five pear varieties (Sydney, Hongxiangsu, Honey pear, Hongxiao pear and Sour pear) with different sizes are collected. The largest pear, Sydney has an average equatorial circumference of 27.64 cm and a weight of 362.84 g. The smallest pear has an average equatorial circumference of 18.35 cm and a weight of 112.67 g. A total of 197 samples. The spectral range is 900~1 700 cm-1, three points were selected on the equator of the pear to measure the three chemical indexes of the pear fruit: soluble solids, acidity (PH) and hardness. It was found that the absorbance of small pear was higher than that of large pears. The three-point average spectrum is used to represent the spectrum of the sample and the first-order derivative pretreatment, which improves the consistency of the spectrum, and solves the influence of factors such as sample heterogeneity and different pear sizes. The determination coefficients of the correction set of the linear regression model PLS for soluble solids, acidity (pH) and hardness were 0.739 4, 0.933 5, 0.886 6, 0.755 9, 0.873 4, 0.787 4, and 0.550 4, 0.194 1, 0.518 1, respectively. The RMSEP of the prediction set is 0.656 4, 0.242 0 and 0.669 2 respectively. The determination coefficients of the calibration set of the nonlinear regression model LSSVM for soluble solids, acidity (pH) and hardness are 0.976 3, 0.999 9 and 0.996 0 respectively, the determination coefficients of the prediction set are 0.923 4, 0.977 7 and 0.939 4 respectively, and the RMSEC of the calibration set is 0.194 9, 0.003 3 and 0.089 4 respectively. The RMSEP of the prediction set is 0.316 9, 0.108 9 and 0.361 3 in order. Compare linear algorithm with the nonlinear algorithm; the LS-SVM modeling effect is better than PLS. The LS-SVM algorithm ensures that the model is applicable to the prediction of more varieties and a wider quality index range. The accuracy and stability of the model have been significantly improved. It can establish a general model for pears of different varieties and sizes. The handheld near-infrared spectrometer can be used for the rapid, non-destructive and efficient detection of sugar, hardness and pH value of pear fruits, and it has got rid of the limitations of the laboratory. It can realize on-site rapid detection.
Key words：Portable near-infrared spectrometer; Pear; Partial least square; Least squares-support vector machine
毛欣然，夏静静，徐惟馨，韦 芸，陈玥瑶，陈月飞，闵顺耕，熊艳梅. 手持式近红外光谱仪测定梨三种品质指标通用模型建模方法研究[J]. 光谱学与光谱分析, 2024, 44(02): 406-412.
MAO Xin-ran, XIA Jing-jing, XU Wei-xin, WEI Yun, CHEN Yue-yao, CHEN Yue-fei, MIN Shun-geng, XIONG Yan-mei. Study on Modeling Method of General Model for Measuring Three Quality Indexes of Pear by Handheld Near-Infrared Spectrometer. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2024, 44(02): 406-412.
[1] ZHANG Hai-liang, SUN Xu-dong, HAO Yong，et al( (章海亮, 孙旭东, 郝 勇，等). Journal of Northwest A&F University (Nature Science Edition)[西北农林科技大学学报(自然科学版)]，2010, 38(4)：128.
[2] LI Dong-hua, JI Shu-juan(李东华，纪淑娟). Science and Technology of Food Industry(食品工业科技)，2012, 33(21): 312.
[3] Goke A, Serra S, Musacchi S. Hortscience, 2018, 53(5): 669.
[4] LUO Ya-qiong, YU Guo-long, LI Huai-xu, et al(罗亚琼，于国龙，李怀旭, 等). Wireless Internet Technology(无线互联科技), 2021, 18(11): 48.
[5] Lee Y H, Han S H. Bulletin of the Korean Chemical Society, 2016, 37(2): 166.
[6] Yuan L M, Mao F, Li L M. Postharvest Biology and Technology, 2020，(160)：111067.
[7] Xia Y, Fan S X, Tian X, et al. Infrared Physics & Technology, 2020，110: 103443.
[8] Li J B, Huang W Q, Zhao C J, et al. Journal of Food Engineering，2013，11(2)：324.
[9] HAN Dong-hai, WANG Jia-hua(韩东海, 王加华). Chinese Journal of Lasers(中国激光), 2008, 35(8): 1123.
[10] LIU Wei, ZHAO Zhong, YUAN Hong-fu, et al(刘 伟, 赵 众, 袁洪福，等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2014, 34(4): 947.
