The data and findings discussed here come from the cited paper and do not represent test results from VISPEK equipment.
When identifying the origin of commercial milk, there is a question that is more worthy of clarification than the recognition rate itself: Is the model identifying differences in origin, or differences in fat content, product type, etc. that occur simultaneously with the origin?
This study gives a very typical case. It demonstrates the potential of near-infrared spectroscopy for commercial milk classification, and also reminds us that in the face of beautiful model results, we still have to go back to the sample design to judge what the model has really learned.
Summary
The study collected near-infrared reflectance spectra of labeled commercial milk from five provinces and compared dimensionality reduction and classification methods. The highest recognition rate reached 97.33%. However, the Hebei group is semi-skimmed and the remaining groups are full-fat. The model may also take advantage of differences in components such as fat. This experiment is suitable for discussing the fingerprint classification and confounding factors of commercial milk, but cannot be used to prove the independent authentication ability of pure geographical origin.
(AI-generated conceptual illustration for context; it is not data from the paper, a photograph of the research equipment, or a VISPEK product diagram.)
Why It Matters
The case of mixed origin and fat labels reminds dairy purchasers that identity libraries must cover different origins of the same ingredients.
Methods and Reference Measurements
The original paper uses a NIR-M-R2 reflection handheld near-infrared spectrometer with a measurement range of 900–1700 nm and an optical resolution of 10 nm (see section 2.3 of the original paper).
There are 60 samples of each of the five types of UHT commercial milk, and 5 mL of each portion is placed in a 10 mL beaker. Keep the same distance from the reflective handheld NIR-M-R2, and average three times between 900-1700 nm. 225 training, 75 testing, SG/SNV and other pre-processing, and PCA dimensionality reduction to compare classifiers such as LDA, DLDA, FDLDA and KNN; the Hebei group is semi-skimmed, and the remaining four groups are full-fat, and the source and fat content are not completely decoupled.
Wavelengths and Acquisition Parameters
| Band/parameter | Original meaning and usage instructions |
|---|---|
| 900-1700 nm | Instrument measuring range; approximately 5882-11111 cm⁻¹ |
| 10 nm | Original optical resolution, 228-dimensional data is not 228 independent narrow bands |
Key Findings
The highest recognition rate of FDLDA-KNN is 97.33%; differences in product ingredients may be used and cannot be used as a pure geographical source to verify performance.
Limitations and Interpretation
The source label of product packaging is not equivalent to the true value of independent ranch; semi-skimmed mixture limits the conclusion of origin.
Image copyright belongs to the paper’s original authors. License: CC BY 4.0.
Source Paper
Wu X, Wang Y, He C, Wu B, Zhang T, Sun J. (2024). Several Feature Extraction Methods Combined with Near-Infrared Spectroscopy for Identifying the Geographical Origins of Milk. Foods (Basel, Switzerland). https://doi.org/10.3390/foods13111783
Contact Us
Whether portable equipment can be used for similar commercial milk testing needs to be evaluated based on samples and experimental conditions, and the specific effect still needs to be verified through actual testing. Welcome to contact VISPEK to exchange your sample and testing needs, jointly evaluate experimental plans, and explore the feasibility of using our company's equipment for actual measurement and verification.
- Email: contact@vispek.cn
- Website: https://www.vispek.cn
VISPEK products shown for illustration; these were not the devices used in the cited study.