基于二维相关谱不变距特征判别掺杂尿素牛奶

为实现牛奶掺杂尿素的快速客观检测,以浓度为外扰,分别建立纯牛奶和掺杂尿素牛奶的二维相关光谱图库,并采用不变矩统计特征表征所获得纯牛奶及掺杂尿素牛奶的二维相关光谱图;针对提取的二维相关光谱不变矩特征,通过计算其Fisher系数评价类间分离程度,结合主成分分析法进行特征优选,选择4个主成分表征所获得样品的二维相关光谱图特征;将优选的4个特征参数作为输入量,采用支持向量机算法建立掺杂尿素牛奶与纯牛奶间的判别模型,该模型对校正集样品和预测集样品的判别准确率分别为94.4%、84.6%。结果表明,基于二维相关谱不变距特征判别掺杂尿素牛奶是可行的。

Detection of Milk Doped with Urea Based on the Moment Invariants Feature of the 2D Correlation Spectrum

In order to realize the rapid and objective detection of milk doped with urea, the application of two-dimensional correlation spectroscopy for the identification of mixed milk was studied. Based on the concentration of external disturbance, the 2D correlation spectra of pure milk and milk doped with urea were established and the obtained 2D correlation spectra of milk and milk doped with urea was characterized by invariant moments. The 2D correlation spectra invariant moment features was selected based on Fisher coefficient which is used to evaluate the degree of separation between classes and the principal component analysis. These 4 invariant moment parameters were used as input for SVM to build discriminant model of milk doped with urea and pure milk. The recognition rate of calibration set samples and prediction set samples was 94.4%, 84.6%respectively. The results showed that detection method of milk doped with urea based on the moment invariants feature of the 2D correlation spectrum is feasible.