摘要

In order to realize the real-time, fast and accurate testing for the main ingredients of milk, the paper discusses the analytical methods which measures the main ingredients of milk by the near infrared diffuse reflectance. In the process of experiment, eigenvalues are extracted based on the near infrared characteristics of different ingredients for milk, the prediction models for protein, fat and lactose contents are established by multiple linear regression, and the samples out of the model are forecasted, the average relative error values for protein, fat and lactose are 2.32%, 3.05% and 4.58% respectively among the results. The experimental results show that, using the appropriate data preprocessing methods and feature extraction ways, the protein, fat and lactose contents of milk can be forecasted accurately by near infrared diffuse reflectance spectroscopy technology.

  • 出版日期2013

全文