论文部分内容阅读
为探索应用近红外光谱技术检测玉米单籽粒蛋白质含量,本研究采用JDSU近红外光谱检测仪采集了205份不同基因型玉米材料的单籽粒光谱值,用常规化学法测定玉米单籽粒蛋白质含量化学值,以117个样本为建模集,拟合了玉米单籽粒近红外光谱仪扫描得到的光谱图与玉米单籽粒蛋白质含量化学值之间的相互关系,用88个样本作预测集,比较了偏最小二乘回归法(PLSR)和支持向量机回归法(SVR)2种预测模型的效果。结果表明,玉米单籽粒种子的蛋白质含量在样本中变异范围为3.48%~18.15%,平均值为10.17%。偏最小二乘回归法(PLSR)和支持向量机回归法(SVR)所建的模型预测效果基本相同,其决定系数(R2)分别为0.99和0.99,校正标准差(SEC)分别为0.32和0.32,预测标准差(SEP)分别为0.46和0.46,相对预测标准差(RSEP)分别为4.61和4.60,RPD分别为6.106和6.111。上述参数表明PLSR和SVR所建立的模型预测效果都比较好,预测值基本接近参比值,便携式JDSU近红外光谱检测仪可以应用于定量分析玉米单籽粒蛋白质含量。
In order to explore the application of near infrared spectroscopy to detect single grain protein content in corn, this study used JDSU near infrared spectroscopy instrument to collect single grain spectral data of 205 different genotypes of maize. The chemical value of single grain protein content , A set of 117 samples was used to fit the correlation between the spectra obtained by single grain near infrared spectrometer and the chemical values of protein content of single grain of corn. 88 samples were used as the predictive set to compare the partial least squares The results of PLSR and support vector machine regression (SVR) two prediction models. The results showed that the variation of protein content in single kernel seeds was 3.48% ~ 18.15% with the average value of 10.17%. The PLSR and SVR models have the same predictive effect, with R2 (0.99 and 0.99) and the standard deviation of correction (SEC) of 0.32 and 0.32, respectively , Respectively. The predicted standard deviations (SEP) were 0.46 and 0.46 respectively. The relative standard deviation of prediction (RSEP) were 4.61 and 4.60 respectively, and the RPD was 6.106 and 6.111 respectively. The above parameters indicate that the model predictions made by PLSR and SVR are good and the predicted values are close to the reference values. The portable JDSU NIR detector can be used to quantitatively analyze the grain protein content of corn.