扫描次数对椰子肉含油量近红外光谱建模的影响

Influences of Scanning Times on the Modeling of Oil Content of Coconut Meat with Near Infrared Spectrum(NIR)

分析扫描次数对椰肉含油量近红外光谱建模的影响,为建立性能稳定的近红外定量模型选择合适参数提供参考。采用国产光栅近红外光谱仪,对38份成熟椰肉样本进行近红外光谱扫描,同一样品重装2次,扫描次数分别为3、8、16、32和64次,比较不同扫描次数条件下建立的模型差异,并对模型外部验证集的样品含油量进行预测值和实际值的成对t检验,比较不同模型的预测能力。结果显示,扫描次数为64次时,模型稳定性最好,此时含油量检测模型相关系数为0.8776;成对t检验表明,扫描3~64次条件下所建立的模型,其预测值与真实值间无显著差异。同一样品扫描64次所建立的模型稳定性最好,能够对样品进行准确的预测,但在大批量样品建模的条件下,采用3次扫描的参数,可以提高工作效率。

To analyze the effect of scanning times on the NIR spectral modeling of coconut oil content and select appropriate parameters for the establishment of a stable NIR quantitative model.A homemade NIR grating diffuse spectrometer was used to scan 38 samples of mature coconut meat.The same sample was reloaded twice,and scanned 3 times,8 times,16 times,32 times and 64 times,respectively.The difference of models under different scanning times was compared,and Paired-samples T test of the predicted and actual oil content of the samples from the external validation set of the models were performed to analyze the prediction ability of the models.When the samples were scanned 64 times,the model had the best stability,and the correlation coefficient of the oil content detection model is 0.8776.Paired-samples T test showed that there is no significant difference between the predicted and the true values of the models established with the data from the samples scanned 3 to 64 times.The model established from the data of the coconut meat samples scanned 64 times has the highest stability and can predict the oil content of the samples accurately.However,for modelling of mass samples it is recommended to adopt the data of the samples scanned 3 times,which can improve the work efficiency.