The data and findings discussed here come from the cited paper and do not represent test results from VISPEK equipment.
Screening the source of green coffee beans can start with spectral fingerprints, but samples from the same year and the same processing method cannot answer all purchasing scenarios. This study compares washed Arabica green beans from different sources and demonstrates the classification effect of preprocessing and neural networks, while also retaining reminders of possible confounding by origin and batch.
Summary
The study collected reflection spectra of four washed Arabica green beans from Indonesia, combined with preprocessing, principal component dimensionality reduction and neural network classification, and achieved high recognition performance in the research test. The evidence comes from market samples of the same year, and has not been verified in the new harvest season or from unknown sources; there are also differences in the scope of the original text, and the conclusion cannot be directly transferred to roasted beans.
(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 double range comparison of the same green beans provides a reference for the selection and configuration of the coffee purchasing library. The follow-up focus is on cross-year and cross-processing batches.
Methods and Reference Measurements
The study collected washed Arabica green beans from four Indonesian sources in 2022, 600 beans from each source, and measured them using two reflectance spectrometers under light-shielding conditions. Compare the original spectrum, smoothing, scattering correction and first derivative, first undergo dimensionality reduction by PCA and then input into the multi-layer perceptron. The samples were randomly divided into two-thirds for training and one-third for testing. The training set was then subjected to ten-fold cross-validation; the classification indicators of the two spectral ranges and different preprocessing were compared. Different spectral preprocessings were compared independently, without considering origin as chemical composition.
Wavelengths and Acquisition Parameters
| Band/parameter | Original meaning and usage instructions |
|---|---|
| Methods 400–1000 and 970–1630 nm; results are described separately SWNIR954–1700 nm, inconsistent range needs to be verified | Collection/analysis scope of this study |
| Discussion of 1210, 1300, 1450, 1570 nm component-related regions; not a small number of verified channel combinations | Original text features and evidence description |
Key Findings
Table 6 of the original text shows that the average accuracy of the near-infrared MSC test is 99.81%; the SNV 10-fold verification is 99.25±0.54%. Indicators are measured according to the average classification of the original text, and have not been verified for new harvest seasons or unknown sources.
Limitations and Interpretation
Method and result SWNIR ranges are inconsistent; processing methods and origin/batch differences may be confounding and cannot be transferred to roasted beans.
Image copyright belongs to the paper’s original authors. License: CC BY 4.0.
Source Paper
Dharmawan A, Masithoh RE, Amanah HZ. (2023). Development of PCA-MLP Model Based on Visible and Shortwave Near Infrared Spectroscopy for Authenticating Arabica Coffee Origins. Foods. https://doi.org/10.3390/foods12112112
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.