基于高光谱图像技术的配合饲料主要营养成分的检测方法

发布时间：2018-07-20 15:03

【摘要】：收集403个配合饲料样本,利用高光谱成像仪对样本进行图像采集,获取配合饲料样本的可见/近红外光谱信息。采用光谱杠杆值和学生残差法剔除异常样本,利用CG法、SPXY法和K-S法按3∶1的比例进行样本集划分,采用均值中心化、标准化、一阶导数、二阶导数、正交信号校正、多元散射校正和标准正态变量变换、去趋势变换,以及其组合方法对光谱进行预处理;采用相关系数法获取特征波段,建立基于高光谱图像技术的配合饲料中粗蛋白、粗灰分、水分、总磷、钙含量的偏最小二乘法(PLS)定量分析模型。通过验证,粗蛋白验证集决定系数R~2V为0.777 8,均方根误差RMSEP为2.6155%,相对分析误差RPDV为2.114 3;粗灰分验证集R~2V为0.775 8,RMSEP为1.0611%,RPDV为2.120 4;水分验证集R~2V为0.631 4,RMSEP为1.6003%,RPDV为1.937 1,总磷验证集R~2V、RMSEP、RPDV分别为0.467 2、0.1916%、1.357 0;钙验证集R~2V仅为0.440 6,RMSEP为0.1755%,RPDV,为1.310 5。结果表明,所建立的粗蛋白、粗灰分最优定量分析模型预测性能较好,水分最优定量分析模型预测精度不够理想,总磷和钙定量分析模型的预测性能很差。
[Abstract]:403 feed samples were collected and the image was collected by hyperspectral imager to obtain the visible / near infrared spectrum information of the feed samples. The spectral lever value and student residual method were used to eliminate abnormal samples, and CG SPXY method and K-S method were used to divide the sample set according to 3:1 ratio. The mean centralization, standardization, first derivative, second derivative, orthogonal signal correction were adopted. Multivariate scattering correction, standard normal variable transformation, detrend transformation, and its combination method are used to pre-process the spectrum, and the correlation coefficient method is used to obtain the characteristic bands, and the crude protein and ash in the feed based on hyperspectral image technology are established. Partial least square (PLS) quantitative analysis model of water, total phosphorus and calcium content. Through verification, The determination coefficient of crude protein validation set is 0.777 V, the root mean square error (RMSEP) is 2.615555, the relative analysis error is 2.114 3, the crude ash verification set is 0.775 8V RMSEP is 1.0611U RPDV, the water verification set R2V is 0.631 4RMSEP is 1.937 1, and the total phosphorus verification set RMS-EPRPDV is 1.6003V, respectively. The calcium verification set RV is only 0.440 6 RMSEP 0.1755 RPDV is 1.310 5. The results show that the optimum quantitative analysis model of crude ash has good prediction performance, the prediction accuracy of the optimal quantitative analysis model of water content is not ideal, and the prediction performance of the quantitative analysis model of total phosphorus and calcium is very poor.
【作者单位】：华中农业大学工学院;
【基金】：国家公益性行业(农业)科研专项(201003063-04)
【分类号】：S816.8

【相似文献】