A non-destructive testing method for original packaged milk powder based on multi-scale terahertz spectroscopy

Abstract

The invention relates to an original-package milk powder nondestructive testing method based on multi-scale terahertz spectrums. The method includes the steps: acquiring terahertz spectrums of original-package milk powder with known and qualified true values of nutrient components by the aid of a terahertz spectrum signal acquisition system; performing high-density wavelet transformation, and selecting optimal high-density wavelet bases; comparing differences of original spectrum signals between different samples under different decomposition scales, and establishing optimal decomposition scales according to minimum cross prediction errors; performing multi-scale high-density wavelet decomposition on terahertz spectrum signals to obtain a high-density wavelet coefficient of each layer; building a predictive sub-model for the obtained high-density wavelet coefficient of each layer; building corresponding predictive sub-models for the high-density wavelet coefficients of different layers to form a predictive sub-model group; fusing all predictive sub-models through suitable fusion strategies.

Description

technical field [0001] A non-destructive testing method for original packaging milk powder based on multi-scale terahertz spectroscopy Background technique [0002] Milk powder has the advantages of rich nutrition, reasonable composition ratio, and easy absorption by the human body. It is loved by people and has long been the main source of nutrition for infants and one of the main nutritional supplements for middle-aged and elderly people. However, driven by high profits in recent years, milk powder safety issues such as "melamine" and "leather milk" dominated by fraudulent nutritional ingredients have emerged one after another, which has become one of the main contradictions in my country's political, economic, and public health fields. . However, most of the traditional milk powder detection methods rely on high-value and easy-to-consume instruments and equipment, which have problems such as high detection cost, cumbersome sample pretreatment, and the need to unpack and t...

AI Technical Summary

Problems solved by technology

However, in the preprocessing process, different preprocessing algorithms have great differences in the emphasis on overcoming spectral interference, which may easily cause the loss of effective information, making the prediction effect and robustness of the built model poor.

In the face of terahertz spectra of complex systems, the shortcomings of this type of algorithm are particularly significant, because the spectral peaks of terahertz spectra are relatively wide, and simple spectral preprocessing is difficult to accurately extract the information of terahertz spectral bands of substances to be measured

Images