不同分集方法对牛肉嫩度高光谱检测模型的比较

Comparative research on hyperspectral detection model of beef tenderness with different sample set partitioning methods

选用牛肉嫩度作为研究对象,开展了4种不同样品集划分方法的选取对其高光谱模型的影响研究。首先选取了70个具有代表性的牛肉样品并提取其肌肉感兴趣区域(ROIs)的光谱,比较分析了浓度梯度法(C-G)、随机法(R-S)、Kennard-Stone(K-S)和光谱-理化值共生矩阵法(SPXY)获取的校正集建立的牛肉嫩度PCR和PLSR模型效果。结果表明:在PCR和PLSR中,SPXY均为最适的样品分集方法,并且4种样品集划分方法下的PLSR模型效果均较优。最优模型SPXY-PLSR校正集的相关系数(Rcal)和均方根误差(RMSEC)分别为0.94和0.48,预测集的相关系数(Rp)和均方根误差(RMSEP)分别为0.93和0.63。研究表明SPXY方法结合高光谱PLSR模型能够实现牛肉嫩度的快速无损检测。

Selection of sample set partitioning methods has an influence on the prediction results of hyperspectral quantitative analysis model. In this paper,beef tenderness was chosen as the research object,and study on the effects of 4 different sample set partitioning methods on hyperspectral model was carried out. Initially,70 representative beef samples were adopted and spectra of muscle regions of interest（ ROIs） were extracted. Subsequently,PCR and PLSR model effects of beef tenderness were separately compared and analyzed,whose models were established by the calibration set separately obtained from concentration gradient method（ C-G）,Random Sampling（ R-S）,Kennard stone（ K-S） and sample set partitioning based on joint X-Y distance（ SPXY）. The results indicated SPXY was the best sample set partitioning method in PCR and PLSR models,the PLSR model effects under 4 sample set partitioning methods were all better than that of PCR. And SPXY-PLSR model effect was optimal,which correlation coefficient（ Rcal） and root mean square error（ RMSEC） of calibration set were 0. 94 and 0. 48 respectively,correlation coefficient（ Rp） and root mean square error（ RMSEP） of prediction set were 0. 93 and 0. 63 respectively. The research shows that the SPXY method combined with hyperspectral PLSR model can realize the rapid and nondestructive detection of beef tenderness.