近红外光谱技术快速鉴别原料肉掺假的可行性研究

探讨利用近红外光谱技术结合Fisher两类判别法以及多层感知器(multilayer perceptron,MLP)神经网络快速无损鉴别原料肉是否掺假,并建立多种掺假肉的分类识别模型的可行性.首先近红外结合主成分与Fisher两类判别,建立原料肉与掺假肉的判别函数,以原料肉与注水肉两类样木的平均重心即两类样木的加权平均...

A Feasibility Study of Rapid Discrimination of Raw Meat and Adulterated Meat Based on Near-Infrared Spectroscopy and Artificial Neural Net Work Model

Feasibility of a nondestructive detection method based on near infrared reflectance （NIR） spectroscopy and chemometrics was put forward for discriminating adulterated meat added with other materials and establishing the classifying recognition model for adulterated meats. First, near-infrared combination of principal components and Fisher were set up to discriminate raw meat and adulterated meat, and the value -0.657 for the weighted mean was set as the distinction threshold. The result indicated that 2/20 samples was wrongly judged and the distinguishing rate was 90%. Then, near-infrared combined principal components and MLP neural network were used to establish three layer neural network identification model for raw meat and three types of adulterated meats, and the recognition rate was 94.2% in the model prediction set consisting of 52 samples. Accordingly, NIRS near-infrared method combined with chemometrics has the potential to detect adulteration in raw meats and to recognize the adulteration categories.