基于近红外光谱的玉米籽粒CNCPS组分分析及预测研究

Analysis and Estimate of Corn CNCPS Component by Near Infrared Reflectance(NIR) Spectroscopy

试验旨在研究应用近红外光谱技术快速测定玉米籽粒粉末CNCPS组分的可行性。65个样品来自黑龙江省,选用偏最小二乘法(PLS)为建模方法,采用二阶导数和Norris导数滤波法处理光谱数据后,建立了玉米籽粒粉末中干物质(DM)、粗蛋白质(CP)、粗脂肪(Fat)、粗灰分(Ash)、淀粉(Starch)、中性洗涤纤维(NDF)、酸性洗涤纤维(ADF)、可溶性蛋白(SP)、酸性洗涤不溶蛋白(ADIP)和中性洗涤不溶蛋白(NDIP)等的近红外预测模型。其中DM,CP,Fat,Ash,Starch,NDF和ADF的决定系数分别为0.974 3,0.968 3,0.947 8,0.909 8,0.977 7,0.935 4和0.926 9,标准差(SD)与预测均方根(RMSEP)的比值(SD/RMSEP)值分别为3.96,4.78,3.75,4.25,4.13,3.88和3.12。SP的决定系数为0.857 5,SD/RMSEP值为3.06。ADIP和NDIP的决定系数分别为0.531 9和0.683 3,SD/RMSEP值分别为5.50和2.85。试验结果表明,近红外技术可以用于玉米籽粒粉末CNCPS组分的快速测定,但降低ADIP和NDIP测定误差有待进一步研究。

The objective of the present study was to investigate the feasibility of predicting the CNCPS （eornell net carbohydrate and protein system） composition of corn by near infrared reflectance spectroscopy （NIRS）. Sixty-five corn samples from Hei- longjiang province were used. The partial least square （PLS） regression method, second derivative and Norris derivative filter were applied in the NIRS prediction of CNCPS. For dry matter, crude protein, ash, fat, starch, neutral-detergent fiber and acid- detergent fiber, the determination coefficients were 0. 974 3, 0. 968 3, 0. 947 8, 0. 909 8 ,0. 977 7, 0. 935 4 and 0. 926 9, and the SD/RMSEP values for them were 3.96, 4. 78, 3.75, 4. 25, 4. 13, 3.88 and 3.12, respectively. The determination coeffi- cient and SD/RMSEP value were 0. 857 5 and 3.06 for soluble protein, but low determination coefficients of 0. 531 9 and 0. 683 3 with SD/RMSEP values of 5.50 and 2.85 were observed for acid-detergent insoluble protein and neutral-detergent insoluble pro- teirL If the SD/RMSEP value 〈5 and 〉3, then the effect of model is ideal, and if the SD/RMSEP value 〉5 or 〉3, the effect of model is not ideal, and at this time, the degree of accuracy of model needs further to be improved. The results of this study indicated that corn nutritive values could be fast and accurately predicted by NIRS. This model was significant in practice for en- riching the rapid quantitative methods of determining animal feed materials.