采后葡萄可溶性固形物含量的高光谱成像检测研究

提出一种应用高光谱成像技术检测葡萄可溶性固形物含量的方法。使用高光谱成像系统采集葡萄漫反射光谱,在500~1 000 nm光谱,采用多元散射校正(MSC)、标准正态变换(SNV)进行光程校正,结合一阶微分(1-Der)、二阶微分(2-Der)、Savitzky-Golay(S-G)平滑方法及其组合对原始光谱进行预处理,建立可溶性固形物含量的偏最小二乘法(PLS)和逐步多元线性回归(SMLR)模型。结果表明:采用PLS和SMLR建模方法均取得较好的预测效果。采用经过MSC、1-Der和S-G平滑相结合预处理后的光谱建立PLS预测模型,校正集的相关系数Rc为0.979 1,RMSEC为0.265,预测集的相关系数Rp为0.962 0,RMSEP为0.372;采用原始光谱、1-Der和SG平滑相结合预处理后的光谱建立SMLR预测模型,校正集的相关系数Rc为0.967 8,RMSEC为0.327,预测集的相关系数Rp为0.947 2,RMSEP为0.394。以上表明,基于高光谱成像技术可以实现采后葡萄可溶性固形物含量的准确无损检测。

Detection of Soluble Solids Content of Postharvest Grape Based on Hyperspectral Imaging

A detection method of soluble solids content in grape based on hyperspectral imaging technology was proposed.The hyperspectral imaging system was used to collect the diffuse reflectance spectra of grape.In the spectra of 500-1 000 nm,multiplicative signal correction(MSC) and standard normal variate(SNV) were used to correct spectra,combined with first derivative(1-Der),second derivative(2-Der),Savitzky-Golay(S-G) smoothing and their combinations to preprocess the original reflectance spectra.Different calibration models of soluble solids content were developed based on partial least square(PLS) and stepwise multiple linear regression(SMLR).The results showed that the modeling effects were both very good using the method of PLS and SMLR.PLS prediction model was established after the pretreatment of MSC,1-Der and S-G smoothing.The correlation coefficient of calibration set(Rc) was 0.979 1 with the root mean square error of calibration set(RMSEC) of 0.265,and the correlation coefficient of prediction set(Rp) was 0.962 0 with root mean square error of predition set(RMSEP) of 0.372.SMLR prediction model was established after the pretreatment of original spectra,1-Der and S-G smoothing.The Rc was 0.967 8 with RMSEC of 0.327,and the Rp was 0.947 2 with RMSEP of 0.394.The study showed that hyperspectral imaging technique could determine soluble solids content of postharvest grape accurately and non-destructively.