基于近红外光谱的腐乳白坯硬度检测研究

Detection of Tofu Hardness by Near Infrared Spectroscopy

考察了硬度与腐乳白坯中水分含量和蛋白质含量的相关关系,探讨了利用近红外光谱技术检测白坯硬度的可行性。通过水分以及蛋白质的相关吸收峰建立预测白坯硬度的数学模型;在建模过程中重点讨论了多元散射校正(MSC)、一阶求导和波段选择等优化处理对建模的影响,利用偏最小二乘法得到的最优模型的建模相关系数r=0.935,建模标准差RMSEC=0.019 3,预测标准差RMSEP=0.023 6,其分级正确率达到94.72%;利用主成分分析法结合判别分析法建立的定性判别模型,分级正确率也达到了90.12%。上述分级结果均好于感观评价的方法,表明近红外技术可以实现白坯硬度的快速无损检测。

The possibility of determination of tofu hardness by near infrared spectroscopy （NIR, 833-2500 nm） was studied. The influences of water content and protein content on the spectra were discussed, in order to detect the hardness. The models of Tofu hardness were calibrated by partial least squares regression （PLS） after eliminating outline, spectra pretreatment, and wavelength optimization （r=0. 994, RMSEC=0. 391M, RMSEP= 0. 416%, RSD= 2.27M）. All of the results were better than those by sense assessing method.