基于反射光谱特征的牛肉嫩度预测模型研究

Prediction model of beef tenderness based on VIS/NIR reflectance spectral characteristic

旨在研究牛肉反射光谱在不同波段范围内的特征,筛选与牛肉剪切力值相关性大的敏感波段,优选牛肉剪切力值的预测模型。实验采用双通道光谱仪采集72块牛肉样品的反射光谱曲线,对原始光谱及剪切力值进行相关性分析,应用联合区间偏最小二乘法（Si-PLS）和遗传算法（GA）筛选敏感波段,建立牛肉剪切力值的偏最小二乘回归（PLSR）预测模型,并对模型进行了优选。结果表明：实验所用2个波段光谱均能实现对牛肉剪切力值的预测,在（400~1000）nm波段建立的预测模型结果较好,在可见光范围内光谱信号与剪切力值呈较大的正相关性,当联合9个子区间作为输入变量时建立PLSR模型的预测相关系数R和均方根误差分别达到0.8598和6.0141,模型的RPD值为1.91。应用遗传区间偏最小二乘法（GA-i PLS）在已选的9个子区间内再次筛选变量,所建立模型的验证集相关系数为0.8883,均方根误差为5.5665,RPD值达到了2.06。研究结果表明,可见近红外反射光谱可以实现对牛肉嫩度的较好预测,基于特征波段的预测模型可以为牛肉嫩度在线快速检测提供理论依据。

The aim of this article was to explore the differences of beef spectrum signal in different wave range, to select sensitive wavelengths associated with beef tenderness, to optimize prediction model. Reflectance spectrum of 72 samples from different cattle was collected by two-channel spectrograph in two wave ranges, respectively. Simple correlation values of beef reflectance and Warner-Bratzler shear force value were analysis, also with synergy interval partial least squares（siPLS） and genetic algorithm interval partial least squares（GA-iPLS） were used to select sensitive wave bands, then partial leastsquares regression（PLSR） model of beef tenderness was established. The results showed that both the spectral range could realize the prediction of beef shear force value, based on simple correlation analysis combined siPLS and GA methods, the results indicated that visible region was of great positive correlation with beef shear force value, nine intervals were selected as input variables of PLSR prediction model with correlation coefficient R and root mean square error of 0.8598 and 6.0141, RPD value of 1.91. While using genetic algorithm interval partial least squares（GA-iPLS） within the range of nine intervals to furthermore choose variables, correlation coefficient and root mean square error of the validation set is 0.8883 and 5.5665, RPD value reached 2.06. The study demonstrated that visible and near infrared（VIS/NIR） spectral can predict beef tenderness accurately and rapidly, and the optimization prediction model with selected wave bands can provide theoretical basis for the rapid online detection of beef tenderness.