论文部分内容阅读
可溶性固形物含量(SSC)是评价桑椹鲜果品质的重要指标,利用近红外光谱分析技术建立快速、实时无损地检测桑椹鲜果中可溶性固形物的方法。首先用手持式Micro NIR1700型近红外光谱仪采集桑椹的近红外光谱,对光谱进行预处理后,应用偏最小二乘回归(PLS)法建立桑椹鲜果SSC预测模型,并用随机蛙(Random-frog)和自适应重加权采样(CARS)2种方法筛选出最优波长变量,提高PLS模型预测精度。经过1阶求导(1stDer)、标准正态变量变换(SNV)和均值中心化(MNCN)相结合预处理后的全波长光谱PLS模型的预测效果最好,校正集与验证集的相关系数平方(R2)分别为0.916 1和0.925 0,均方根误差分别为0.985 8°Brix和0.654 3°Brix。相较于Random-frog法,用CARS方法优选出19个波长变量,所建PLS模型的预测效果更好,校正集与验证集的R2分别为0.933 2和0.943 4,均方根误差分别为0.782 0°Brix和0.582 8°Brix。研究结果表明,利用手持式Micro NIR 1700型近红外光谱仪结合化学计量学方法,能够用于现场对桑椹鲜果SCC的快速无损检测。
Soluble solid content (SSC) is an important index to evaluate the quality of fresh mulberry fruit. NIR spectroscopy was used to establish a rapid and real-time non-destructive method for detecting soluble solids in fresh mulberry fruit. Firstly, the near-infrared spectrum of mulberry was collected by hand-held Micro NIR 1700 near-infrared spectrometer. After spectrum preprocessing, the prediction model of fresh mulberry fruit SSC was established by partial least squares regression (PLS) Two methods of Adaptive Weighted Sampling (CARS) were used to select the optimal wavelength variables and improve the prediction accuracy of PLS model. The prediction results of PLS model after first-order derivation (1stDer), SNV and MNCN pretreatment are the best, and the correlation coefficient between calibration set and validation set is squared (R2) were 0.916 1 and 0.925 0, respectively. The root mean square errors were 0.985 8 ° Brix and 0.654 3 ° Brix, respectively. Compared with the Random-frog method, 19 wavelength variables were optimized by CARS method, and the PLS model was better than the random-frog method. The R2 of the calibration set and the validation set were 0.933 2 and 0.943 4 respectively, and the root mean square errors were 0.782 0 ° Brix and 0.582 8 ° Brix. The results show that the use of hand-held Micro NIR 1700 near infrared spectroscopy combined with chemometrics methods, can be used for on-site fast fruit mulberry SCC non-destructive testing.