摘要
本研究利用瑞典波通DA7200型固定光栅连续光谱近红外分析仪对460份来自阿根廷、美国和巴西三个国家的进口大豆加工的豆粕进行了分析，并根据定标样品的吸收光谱与化学分析数据建立了豆粕的水分．蛋白质及脂肪等三项指标的定标模型。定标模型的统计分析结果显示，三项指标的模型均具有较高的相关性（R^2=0．9616-0．9886）．和较低的交互验证标准偏差（SECV=0．0794～0．2160）。选用具有代表性的预测集样品对校准模型的准确性进行预测。预测结果显示：预测标准偏差在0．10～0．27之间，双试样标准差在0．03～0．10之间。近红外光谱技术用于快速测试豆粕品质是可行的，能够用于豆粕生产厂家产品的品质控制。
About 460 soybean meal samples that processing of inported soybean which from Argentina, USA and Brazil were analyzed by DA7200 of Perten instrument in this paper .And three calibration models between absorption spectrum and the data of chemical analysis were established which respectively based on the moisture, protein and fat of soybean meal.According to the statistic results, all of the calibration models obtained satisfied results which have high relativity R2= 0. 9616-0. 9886 and SECt= 0. 0794-0. 2160. Representative samples were also used to verify the model＇s authenticity, The SEP （ 0.10-0.27 ） and SDD （0.030.10） were acquired.The results showed that NIR spectroscopy analysis technology can be used to detect soybean meal rapidly and control the quality of soybean meal for factory.
