基于近红外透射光谱及神经网络的大豆油质量分析  

Quality Analysis of Soybean Oil based on Near Infrared Transmission Spectra and Artificial Neural Network Model

作　　者：蔡立晶[1] 蔡立娟[2] 李文勇[3] 赵肖宇[1] 尚廷义[1] 

机构地区：[1]黑龙江八一农垦大学,黑龙江大庆163319 [2]长春理工大学,长春130022 [3]大庆油田有限责任公司第一采油厂,黑龙江大庆163162

出　　处：《湖北农业科学》2015年第1期175-177,共3页Hubei Agricultural Sciences

基　　金：黑龙江省教育厅科学技术研究项目(12521378)

摘　　要：提出了一种基于近红外透射光谱及最速下降BP算法识别大豆油质量的方法。光谱采集范围是10 000-4 000 cm^-1,将得到的近红外光谱数据作为网络的输入神经元,利用主成分分析方法得出8个变量指标数,该变量指标对样品累计贡献率达到99.9%以上;将8个主成分的特征值作为BP网络的输入向量,建立BP神经网络模型。该模型对预测样品集能正确判别,判别正确率达到100%。A method based on near infrared transmission spectra and gradient descent BP algorithm was used to analyzed the quality of soybean oil. The range of 10 000 to 4 000 cm^-1spectral was acquired, then the near infrared spectrum data was input to BP network. Eight variable indexes were obtained with principal component analysis. The cumulative contribution rate of the 8 variable indexes was more than 99.9%. Using the 8 index as input vectors of BP neural networks model, it can discriminate the quality of samples with the accuracy of 100%.

关 键 词：近红外透射光谱 BP神经网络 豆油质量分析 

分 类 号：O657.33[理学—分析化学]