Handheld NIR Spectroscopy for Non-Destructive Staling Monitoring of Toast Bread: Case 1 – Unsupervised Clustering-Based Algorithms

This study evaluated the potential of a portable visible and near-infrared (Vis-NIR) spectroscopy system, integrated with unsupervised clustering algorithms, for non-destructive characterization of staling in toast bread under different storage temperatures (laboratory and refrigeration) and packaging conditions (packaged and unpackaged). Vis-NIR spectral data were collected from both crumb and crust samples (400–1100 nm) and preprocessed using Savitzky–Golay (SG) smoothing, multiplicative scatter correction (MSC), and standard normal variate (SNV) to reduce noise and scattering. Principal Component Analysis revealed clear variance patterns associated with freshness stages (Fresh, Acceptable, and Stale). Three unsupervised clustering algorithms, Self-Organizing Maps (SOM), K-means, and Artificial Bee Colony (ABC), were applied for the samples' freshness evaluation. Results showed that preprocessing strongly affected data separability: SG performed best for crust kept at room temperature, whereas MSC was more effective for refrigerated samples. Among the clustering algorithms, K-means provided the highest clustering performance with Silhouette indices above 0.70 (0.724 for crust and 0.723 for crumb) and Davies–Bouldin indices below 0.60, indicating well-defined clusters. SOM efficiently visualized internal spectral relationships but showed overlap between acceptable and stale groups. ABC exhibited superior adaptability in determining the optimal number of clusters and achieved the best differentiation for unpackaged refrigerated samples (DB = 0.466; Silhouette = 0.645). Overall, integrating portable NIR spectroscopy with unsupervised learning provides a rapid, accurate, and non-destructive approach for assessing bread freshness. This method can support real-time quality control, minimize waste, and reduce reliance on destructive laboratory analyses in the bakery industry, as well as the distribution and retail chain.