注水羊肉的高光谱数据采集与预处理方法研究

HIGH SPECTRAL DATA ACQUISITION AND PRETREATMENT METHOD OF WATER-INJECTED MUTTON

本文以羊肉为检测对象,采用高光谱成像系统从126个羊肉样本中采集光谱信息。通过去趋势算法、标准正态化、多元散射校正、归一化、基线校准(Baseline)、移动平均平滑(Moving Average)、导数(Derivation)法等多种方法对原始光谱(400~1000 nm)进行预处理。通过建立偏最小二乘法(PlSR)预测模型来选出最优预处理方法。结果表明,经多元散射校正结合标准正态的预处理方法效果最好,预测集相关系数RC为0.7415,标准偏差SEC为0.0624;验证集相关系数RP为0.7384,标准偏差SEP为0.0581。

In this paper,the spectral information of 126 lamb samples was collected through the high light imaging system.The original spectra(400-1000 nm)were pretreated by trend correction,multiple scattering correction,standard normal,normalized calibration,Baseline,Moving Average smoothing,derivative and other methods.The optimal pretreatment method was determined by establishing the prediction model of partial least squares regression(PLSR).The results showed that the pretreatment method with multivariate scattering correction combined with standard normal state had the best performance,the prediction set correlation coefficient Rc was 0.7415,and the standard deviation SEC was 0.0624.The verify set correlation coefficient Rp was 0.7384,and the standard deviation SEP was 0.0581.