基于近红外光谱法建立的定量分析模型快速预测苗药水冬瓜叶中金丝桃苷和异槲皮苷。样品经甲醇水浴回流提取后，采用近红外光谱仪以透射法采集其1 100~2 300 nm波长范围内的光谱信息，同时用高效液相色谱法测定金丝桃苷和异槲皮苷的含量，结果作为参考值。分别以标准正态变量交换法-平滑法（SNV-Smoothing）和一阶微分法（S-G first derivative）预处理金丝桃苷和异槲皮苷的原始光谱，以偏最小二乘法（PLS）分别在1 100~2 300 nm和1 100~1 400 nm波段内构建金丝桃苷和异槲皮苷的定量分析模型，并以相关系数、校正均方根误差（RMSEC）、预测均方根误差（RMSEP）、预测相对偏差（RSEP）来考察模型性能。结果显示：金丝桃苷定量分析模型的RMSEC、RMSEP与RSEP值分别为0.241 4，0.270 5，0.177 5，相关系数为0.913 6；异槲皮苷定量分析模型的RMSEC、RMSEP与RSEP值分别为0.079 1，0.020 8，0.033 9，相关系数为0.907 7；用所建金丝桃苷和异槲皮苷的定量分析模型预测验证集样品中的金丝桃苷和异槲皮苷含量，预测值与参考值的相对误差分别为-2.8%~2.9%、-5.1%~5.9%；对实际样品重复分析6次，金丝桃苷和异槲皮苷的预测值的相对标准偏差分别为2.7%，3.3%；将样品溶液放置0，1，2，3，6，12 h后进样，预测值的RSD分别为2.9%，3.4%.

Quantitative analysis model based on near-infrared spectroscopy was developed for rapid prediction of hyperin and isoquercitrin in Miao medicine of the leaves of Toricellia Angulata Oliv.var.intermedia (Harms) Hu. After the sample was extracted with methanol by reflux in a water bath, the near-infrared spectrometer was used to collect the spectral information in the wavelength range of 1 100-2 300 nm by transmission method. Meanwhile, hyperin and isoquercitrin were determined by HPLC, and the results were used as reference values. The original spectra of hyperin and isoquercitrin were pretreated by the transformation SNV-smoothing and S-G first derivative. The quantitative analysis models of hyperin and isoquercitrin were constructed by partial least squares method (PLS) at 1 100-2 300 nm and 1 100-1 400 nm, respectively. Correlation coefficient, root mean square error of correction (RMSEC), root mean square error of prediction (RMSEP) and relative prediction deviation (RSEP) were used to investigate the performance of the model. The results showed that RMSEC, RMSEP and RSEP values of quantitative analysis model of hyperin were 0.241 4, 0.270 5 and 0.177 5, respectively, and correlation coefficient was 0.913 6. RMSEC, RMSEP and RSEP values of quantitative analysis model of isoquercitrin were 0.079 1, 0.020 8 and 0.033 9, respectively, and correlation coefficient was 0.907 7. The established quantitative analysis models of hyperin and isoquercitrin were used to predict the contents of hyperin and isoquercitrin in the validation set samples. The relative errors between predicted values and reference values were found in the ranges of -2.8%-2.9% and -5.1%-5.9%, respectively. After six repeated analysis of the actual samples, RSDs of the predicted values of hyperin and isoquercitrin were 2.7% and 3.3%, respectively. After placing the sample solution for 0, 1, 2, 3, 6, 12 h, RSDs of the predicted values were 2.9%, 3.4%, respectively.