Firmness is one of the most important characteristics to estimate fruit maturity and quality.The potential of near-infrared(NIR)diffuse reflectance spectroscopy as a nondestructive way for pear firmness evaluation of three varieties(‘Cuiguan’,‘Xueqing’and‘Xizilv’)was studied,both quantitatively and qualitatively.NIR models were established using partial least square(PLS)methods in the spectral range of 800 to 2500 nm.For quantitative analysis,the correlation coefficient r increased with more varieties involved in the model.Best results were obtained in the model for all three varieties:rcalwas 0.934,root mean square error of calibration(RMSEC)and root mean square error of prediction(RMSEP)were 2.06 N and 3.14 N,respectively.For qualitative analysis,the overall accuracies of discriminant PLS models for classifying pears into three firmness levels:low,medium and high firmness level were not so good,percentage of samples correctly classified ranged from 70.63%to 81.25%for calibration and from 56.25%to 74.38%for validation.The results indicate that NIR spectroscopy together with PLS chemometrics method is feasible for quantitative analysis of pear firmness,however,the classification accuracy is too low to put into practical application.
该研究应用近红外（near infrared，NIR）漫反射光谱定量分析技术开展了金华大白桃的糖度检测试验研究。用偏最小二乘回归（partial least square regression，PLSR）方法在800-2500nm光谱范围建模，通过比较果汁和不同部位果肉所对应的相关模型的预测结果发现：用水果3个部位（顶部、中部、底部）共9个检测点的果肉平均光谱和糖度平均值建立的模型的结果比果汁或单独某个部位果肉（3个检测点）所建立的模型的结果要好。在此基础上，分析了光谱微分和散射校正预处理对建模结果的影响，结果显示微分光谱建立的模型不如原始光谱建立的模型的结果好，光谱的散射校正处理（用多元散射校正MSC和标准正态变量变换SNV两种方法）有助于提高模型的预测性能。最终建立桃子果肉平均光谱经MSC和SNV散射校正后与糖度的相关模型，MSC和SNV对建模结果的影响基本一致，MSC-PLSR和SNV-PLSR模型的相关系数Rcal和交互验证相关系数Rcross-v分别为0．997和0．939。该研究表明近红外光谱检测技术可用于金华大白桃糖度的定量分析。
