Pixel-wise assessment of industrial compost transformation by NIR hyperspectral imaging and chemometrics: an early-warning tool for process monitoring
- Peña Rueda, M. , Deryck, A. , Stevens, F. , Fernández Pierna, J.A. , Baeten, V. , Ayora-Cañada, M.J. & Domínguez-Vidal, A. (2027). Pixel-wise assessment of industrial compost transformation by NIR hyperspectral imaging and chemometrics: an early-warning tool for process monitoring. Talanta, 311: 130201.
| Type | Journal Article |
| Year | 2027 |
| Title | Pixel-wise assessment of industrial compost transformation by NIR hyperspectral imaging and chemometrics: an early-warning tool for process monitoring |
| Journal | Talanta |
| Label | U12-0348-Rueda-2027 |
| Volume | 311 |
| Pages | 130201 |
| Abstract | Ensuring reliable monitoring of industrial composting remains challenging due to feedstock heterogeneity and the limited representativeness of point-based spectroscopy. Here, near-infrared hyperspectral imaging (NIR-HSI) combined with multivariate modelling was applied for the first time to characterize an industrial olive mill waste composting process. By sampling across different zones of open windrow piles over two campaigns (spanning 2–59 weeks), partial least squares (PLS) regression models were developed to predict key quality indicators (pH, electrical conductivity, C/N ratio, and organic matter content). The models demonstrated satisfactory performance (Residual Prediction Deviation, RPD = 2.4–3.1). Pixel-wise application of PLS models generated spatially resolved prediction maps, revealing localized differences in degradation dynamics that would likely be overlooked by conventional analytical methods. Furthermore, a pixel-wise PLS-DA strategy was implemented to characterise compositional changes throughout composting. In contrast to the quantitative regression models, this approach identified spectral patterns reflecting the progressive breakdown and transformation of the organic matter, providing insights into compost evolution without the need for reference analysis. The PLS-DA model demonstrated an overall accuracy of 94 %, and the Variable Importance in Projection scores indicated wavelengths associated with nitrogen-containing and aromatic structures as relevant maturation indicators. Classification maps captured the gradual decomposition of the initial substrates and compost homogenization. Calculating class-specific pixels (initial raw materials and finished compost) enabled the construction of relative abundance curves reflecting process evolution. Overall, this NIR-HSI workflow opens new possibilities for compost monitoring reliability and represents a promising basis for future data-driven control strategies in heterogeneous biological processes. |
| Fichier | |
| Lien | https://doi.org/10.1016/j.talanta.2026.130201 |
| Authors | Peña Rueda, M., Deryck, A., Stevens, F., Fernández Pierna, J.A., Baeten, V., Ayora-Cañada, M.J., Domínguez-Vidal, A. |
Team
