摘要本文主要概述了近红外光谱技术的原理、光谱数据的预处理、模型建立、桑叶中克 螨特农药残留定量检测以及近红外光谱技术的发展现状，并对近红外光谱技术在桑叶克 螨特残留定量分析方面的应用研究进行综述。将 0。48g/mL 的克螨特原液用蒸馏水稀释 成浓度梯度 0。005~0。01mg/mL 的 39 个样品，其中 26 个作为校正集，13 个样品作为验证 集。把克螨特样品喷洒在桑叶上后，使用近红外光谱仪直接扫描桑叶。扫描采用透射扫 描，扫描 32 次，每个样品扫描 3 次取平均值。使用 MATLAB 软件对数据进行处理，得到 波长范围在 1350~1600cm-1 的近红外光谱数据为有效数据。使用一阶导数、二阶导数、Mean Center、主成分系数、校正相关系数、预测相关系、内部交叉验证均方根、校正预测均方根等方法对原始光谱数据进行预处理，对比预处理结果发现二阶导数 15 点平滑、MSC 和 MC 三者结合的 方法得到的结果最佳，模型预测相关系数达到了 0。8990，预测均方根 RMSEP 为 0。0123。 将预处理后的数据建模并检验模型，模型的校正标准差为 0。0109952、交叉验证标准差 为 0。0154827 和外部验证标准差为 0。010805。再通过确定主因子数，采用偏最小二乘回 归法将样品预测值与模型进行相关性处理，得到主因子数为 4 时预测值与实际值之间相 关系数最大，达到了 98。25。研究获得的结果表明，近红外光谱技术所建立的模型能够 有效地检测桑叶上的农药残留，并能获得与化学检验较为接近的结果。83574

毕业论文关键词：近红外光谱技术；克螨特；桑叶；农药残留；偏最小二乘回归法

Abstract This paper provides an overview of the near infrared spectroscopy technology principle, spectral data pre processing, model establishment, mulberry leaves in grams and propargite pesticide residual quantitative detection and near infrared spectroscopy technology development status, and of near infrared spectroscopy in the leaves of propargite special residue application of quantitative analysis were reviewed。 The 0。48g/mL propargite solution with distilled water and diluted into 39 sample concentration gradient 0。005~0。01mg/mL, 26 of them as the calibration set, 13 samples as a validation set。 The propargite samples were sprayed on mulberry leaves, mulberry leaves using near infrared spectroscopy scanning。 Scanning was used to scan 32 times, and the average value of each sample was 3 times。 The data were processed by MATLAB software, and the near infrared spectrum data of the wavelength range of 1350~1600cm-1 were obtained。 Using the first derivative, second derivative, mean center, the principal component coefficients, correction correlation coefficient, the prediction phase relations, internal cross validation  RMS, correction prediction mean square root method of the raw spectral data pretreatment, the comparison of pre treatment results show that second-order derivative 15 smooth, combination of MSC and MC method to get the results of best, model predictive correlation coefficient reached 0。8990, root mean square root mean square error of prediction (RMSEP) for 0。0123 prediction。 The pre processed data model and test model, the model's calibration standard deviation is 0。0109952, the cross validation standard deviation is 0。0154827 and the external validation standard deviation is 0。010805。 Again by determining the principal factor number, using partial least squares regression method the sample prediction values and model and correlation processing, was the principal factor number 4 predicted value and the actual value between the maximum correlation coefficient reached 98。25。 The results obtained in this study show that the model established by near infrared spectroscopy technology can effectively detect the mulberry leaf of pesticide residues, and can obtain and chemical examination of relatively close to the results。