花生秧常规营养成分近红外反射光谱预测模型的建立

摘要

为了探究近红外反射光谱 (NIRS) 技术在花生秧常规营养成分预测中应用的可行性, 试验采集河南省花生秧107份, 随机分成定标集 (n=83) 和验证集 (n=24) , 建立花生秧干物质 (DM) 、粗蛋白质 (CP) 、粗脂肪 (EE) 、酸性洗涤纤维 (ADF) 、中性洗涤纤维 (NDF) 、粗灰分 (Ash) 、钙 (Ca) 、磷 (P) 8种常规营养成分含量NIRS预测模型, 并计算验证集预测值与实测值的相关性, 验证模型准确度.结果显示:经过标准正态变换 (SNV) +一阶导数预处理后, 花生秧DM含量预测效果最佳[定标相关系数 (RSQcal) =0.989、决定系数 (RSQv) =0.968 2];经过SNV+去趋势校正+一阶导数预处理后, NDF含量预测效果最佳 (RSQcal=0.966、RSQv=0.937 3) ;经过SNV+去趋势校正+二阶导数处理后, CP含量预测效果最佳 (RSQcal=0.923、RSQv=0.903 6) ;而经过不同预处理后EE、ADF含量的RSQcal> 0.7、0.9>RSQv> 0.7, Ca、P、Ash含量的RSQcal<0.7、RSQv<0.7, 预测效果不理想.综合得出:NIRS技术对花生秧中DM、NDF、CP含量能精准预测, 对EE、ADF含量能粗略预测, 而对Ca、P、Ash含量不能预测.%In order to explore the application feasibility of prediction of routine nutrient of peanut vine by nearinfrared reflectance spectroscopy (NIRS) technology, a total of 107 peanut vine samples in Henan province were collected, the samples were divided into calibration set (n=83) and verification set (n=24) , the 8 kinds of routine nutrient contents in peanut vine were analyzed included:dry matter (DM) , ether extract (EE) , crude protein (CP) , acid detergent fiber (ADF) , neutral detergent fiber (NDF) , crude ash (Ash) , calcium (Ca) and phosphorus (P) , and calculated the correlation between predicted value and measured value of verification set, verified the accuracy of the model.The results showed that after pre-treatment of standard normal variate (SNV) + 1st derivative, the predictive effect of DM content in peanut vine was the best [calibration correlation coefficient (RSQcal) =0.989, coefficient of determination (RSQv) =0.968 2];after pre-treatment of SNV+detrended correction + 1st derivative, the predictive effect of NDF content was the best (RSQcal=0.966, RSQv=0.937 3) ;after pre-treatment of SNV + detrended correction + 2nd derivative, the predictive effect of CP content was the best (RSQcal=0.923, RSQv=0.903 6) ;after different pre-treatment, the RSQcal of ADF and EE contents were> 0.7, and 0.9>RSQv> 0.7;and the RSQcal of Ca, P and Ash contents were<0.7, RSQv<0.7, the predictive effect was not satisfactory.It is concluded that the contents of DM, NDF and CP in peanut vine can be accurately predicted by NIRS technology, the contents of EE and ADF can be predicted roughly, while the contents of Ca, P and Ash cannot be predicted.