利用可见-近红外光谱对在4℃下冷藏24 h的灵武长枣脆度进行检测,并建立了最优模型.通过400～1 000 nm高光谱成像系统采集了112个长枣图像,对原始光谱与经SNV,MSC、S-G、1ST、2ND、SNV+ 1ST、MSC+ 1ST、SNV+ 2ND、MSC+ 2ND、SNV+ S-G、MSC+ S-G预处理后光谱的偏最小二乘回归(PLSR)模型进行了对比分析;采用主成分分析法(PCA)、连续投影算法(SPA)、竞争性自适应重加权法(CARS)提取特征波长,分别建立偏最小二乘回归(PLSR)和主成分回归(PCR)模型;将经预处理后的简化PLSR模型与全波段PLSR模型进行了对比分析.结果表明,采用标准归一化法(SNV)预处理后的PLSR模型优于原光谱及其他预处理方法;提取特征波长后建立的CARS-PLSR模型优于CARS-PCR模型和全波段PLSR模型,其相关系数(RP)和预测均方根误差(RMSEP)分别为0.919、1.121.这表明,基于可见-近红外光谱检测冷藏灵武长枣脆度是可行的,SNV-CARS-PLSR模型最佳.

A model of predicting the brittleness of Lingwu jujube stored at 4 ℃ for 24 hours by Visible-near infrared Spectroscopy was established and optimized.Hyperspectral images of 112 jujubes of samples over 400-1000 nm were acquired.PLSR(Partial least squares regression,PLSR)models of raw spectral and spectral processed by SNV,MSC,S-G,1ST,2ND,SNV +1ST,MSC +1ST,SNV +2ND,MSC +2ND,SNV +S-G,MSC +S-G for brittleness were compared.Characteristic wavelengths were selected by principal component analysis (PCA),successive projections algorithm (SPA) and competitive adaptive reweighted sampling (CARS);PLSR models and PCR models on characterizing wavelengths were established.The simplified PLSR model after pretreatment and completed full wavelength PLSR model were compared.The results showed that the PLSR model with pretreatment by SNV was superior to CARS-PCR and full spectralspectral pretreatment.The optimal wavelengths by CARS-PLSR model had an excellent ability to predict the brittleness of jujube and was better than CARS-PCR model and PLSR model to predict brittleness composition.The correlation coefficient (Rp) and root mean square error of prediction (RMSEP) were 0.786 and 1.224 respectively.Therefore,it' s possible to determine the brittleness of chilled Lingwu jujubes by Visible-near infrared Spectroscopy and the SNV-CARS-PLSR model was the best.