基于支持向量机的牛肉嫩度等级评价模型研究

Research on support vector machine evaluation model of beef tenderness

中文关键词:  反射光谱  支持向量机  嫩度  主成分分析

英文关键词:reflectance spectrum  support vector machine  tenderness  principal component analysis

中文摘要:

      目的 利用VIS/NIR反射光谱建立基于支持向量机的生鲜牛肉嫩度等级的评价模型。方法 以牛肉背最长肌为研究对象, 选取样本58个, 在牛肉新鲜切口处采集波长范围400~1700 nm的反射光谱信息, 使用肉类嫩度测量仪测量牛肉剪切力值, 应用支持向量机(SVM)模型评价牛肉嫩度等级。结果 应用SVM模型可以较好地实现对牛肉嫩度等级的评价。尤其是经主成分分析降维预处理, 结合径向基核函数SVM, 对牛肉训练集嫩度等级的回判率达到95%, 对样品校正集判别的准确率进一步提高至83.3%。结论 SVM模型对牛肉嫩度等级评价结果较好, 进行主成分分析后, 判别结果有所提高。

英文摘要:

      Objective To establish a support vector machine evaluation model of beef tenderness using VIS/NIR reflectance spectroscopy. Method A total of 58 strip loins samples were collected from longissimus of beef. The reflectance spectra range of 400~1700 nm were extrcted from fresh cut of beef, and then Warn-er-Bratzlar shear force (WBSF) values were measured. Evaluation model of beef tenderness was established using support vector machine (SVM) model. Results Application of SVM model showed good result for beef tenderness grade evaluation. Especially, using principal component analysis (PCA) after dimension reduction pretreatment, combined with radial basis function (RBF) of SVM, the accuracy researched 95% in training set, and 83.3% in calibration set. Conclusion The SVM model with PCA pretreatment demonstrated great performance for beef tenderness grade evaluation.