为了改善近红外(NIR)光谱测量精度，提出了以中等浓度样品作为参考物的测量方案。从理论上分析了扣除参考背景后吸光度的误差来源，发现以中等浓度样品作为参考物有可能减小样品池和光纤损耗导致的测量误差。以葡萄糖为分析对象，开展了葡萄糖水溶液和3% Intralipid 仿体溶液的近红外透射验证实验。在糖水溶液的双光路实验中，以中等糖浓度样品作为参考物建立的偏最小二乘(PLS)模型的校正均方根误差(f_RMSEC)和交互验证均方根误差(f_RMSECV)比纯水背景分别下降了36.6%和41.0%；而仿体实验中，以中等糖浓度样品作为参考物时，模型的校正均方根误差和交互验证均方根误差比无糖3% Intralipid 溶液作为参考物时分别下降了66.3%和39.6%。实验结果表明以中等浓度样品作为参考物能有效提高近红外透射光谱的分析精度。

In order to improve the accuracy of near-infrared (NIR) spectroscopy measurement, the method with medium concentration sample as reference for relative measurement is proposed. Firstly, theoretical analysis of absorbance error sources with background reducted is presented, which shows that errors caused by sample cell and optical fiber loss can be reduced by using medium concentration sample as reference. Furthermore, NIR transmission experiments on glucose aqueous solution and 3% intralipid solution as phantom are performed with glucose used as analyte. In the double-beam experiment on glucose aqueous solution, root mean square error of calibration (f_RMSEC) and root mean square error of cross validation (f_RMSECV) of the partial least squares (PLS) regression model using medium concentration sample as reference decrease respectively by 36.6% and 41.0% compared to the model with pure water as reference. Meanwhile, in the intralipid phantom experiment, fRMSEC and fRMSECV of the PLS regression model using medium concentration sample as reference decrease respectively by 66.3% and 39.3% compared to the model with glucose- free 3% intralipid solution as reference. The results show that the accuracy of NIR transmission spectroscopy analysis can be effectively improved by using medium concentration sample as reference.concentration sample