The data and findings discussed here come from the cited paper and do not represent test results from VISPEK equipment.
Whether the spectra of different brands of commercial milk can be distinguished is a very intuitive issue when establishing a dairy fingerprint database. However, brands often come with recipe and production batch changes. To determine what the model has learned, you need to look at both the source and the way the sample is divided.
Summary
The study collected portable near-infrared spectra of five brands of milk and completed the classification after preprocessing, dimensionality reduction and discriminant modeling. The fuzzy improvement method performs better in the same batch of test samples, but each brand is only included in one production batch, and the results cannot be directly used as cross-date brand certification capabilities, nor do they represent nutritional or quality rankings.
(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 brand spectral library can be used for batch fingerprint research, and the actual purchase value can be judged only after adding samples across dates.
Methods and Reference Measurements
The original paper used a NIR-M-R2 near-infrared spectrometer to collect milk sample spectra with a measurement range of 900–1700 nm and a resolution of 10 nm (see section 2.2 of the original paper).
The study purchased five brands of milk from local supermarkets, 60 servings of each brand, a total of 300 servings; samples of the same brand came from the same production batch. Three near-infrared spectra of each sample were collected at room temperature and averaged. After smoothing noise reduction and principal component dimensionality reduction, traditional linear discrimination, improved zero space discrimination and fuzzy improvement methods were compared. 45 copies of each brand were used for training and 15 for testing. The recognition results were then evaluated by the nearest neighbor classifier to examine the impact of the number of features and neighbor settings. The original study did not include samples of the same brand with different production dates.
Wavelengths and Acquisition Parameters
| Band/parameter | Original meaning and usage instructions |
|---|---|
| 900–1700 nm; resolution 10 nm; 228-dimensional data | Collection/analysis scope of this study |
| The paper uses dimensionality reduction and discriminant features, and has not verified the optimal small number of discrete wavelength combinations. | Critical wavelengths and evidence boundaries |
Key Findings
Among the 75 test samples from the same batch, FiNLDA has the highest accuracy of 94.67%, compared with 74.7% for LDA and 88% for iNLDA; it is not a cross-date brand certification.
Limitations and Interpretation
Brands and production batches may be mixed; classification does not equal nutritional quality, nor does it prove that well-known brands are of higher quality.
Image copyright belongs to the paper’s original authors. License: CC BY 4.0.
Source Paper
Wu X, Fang Y, Wu B, Liu M. (2023). Application of Near-Infrared Spectroscopy and Fuzzy Improved Null Linear Discriminant Analysis for Rapid Discrimination of Milk Brands. Foods. https://doi.org/10.3390/foods12213929
Contact Us
Whether portable equipment can be used for similar detection 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.