背景和光程对茶多酚含量近红外预测模型的影响
Effects of background and optical path length on predicting the contents of tea polyphenol with NIR models

为比较空气和纯水背景的模型效果，扫描5个（2、3、5、8和10 mm）不同光程的红茶汤近红外光谱，应用排列组合原理和偏最小二乘法(PLS)建立2种背景下单光程和多光程组合茶多酚近红外预测模型，并用t-test检验模型预测效果。结果表明:单光程模型中，纯水背景与空气背景模型预测结果间无显著性差异；多光程组合模型中,空气背景模型要全部优于纯水背景模型，并且最佳组合模型（（5+8） mm）具有比单光程模型更好的预测精度。空气背景茶多酚模型要优于纯水背景茶多酚模型。
In order to establish the stable models,the air and pure water background NIRS models were compared.2 mm,3 mm,5 mm,8 mm and 10 mm optical path length NIR spectra of black tea’s solution on air and pure water background were attained to set up single-optical and multi-optical path length NIR models through using partial least squares (PLS) and principle of permutation and combination.The results of using NIR models to predict unknown samples of black tea were examined by t-test statistical method.The results showed that there was no difference between the NIR models with the same optical path length and different background.There was significant difference (P＜0.05) between models with different optical path lengths on the pure water background.In the multi-optical path length models,the performances of NIR models on air background were better than that of the models on water background.The best NIR model combined 5 mm and 8 mm optical path length,whose predictable accuracy was better than that of single-optical path length models.The NIRS models of air background was better than that of pure water background.