Intelligent Detection of Mold Infection Stage in Cherries Based on Hyperspectral Imaging ( HSI )

This study addresses the critical challenge of early detection for Alternaria alternata infection in postharvest cherries by developing a hyperspectral imaging (HSI) system integrated with machine learning (ML) algorithms. After acquiring HSI data from cherries at different infection stages, principal component analysis (PCA) was first employed to demonstrate the superior discriminative capability of HSI‐based multivariate analysis over conventional color imaging. We then systematically evaluated the effects of preprocessing methods (raw data [RAW] vs. first‐derivative [FD] transformation) and wavelength selection algorithms (Successive Projections Algorithm [SPA] and Competitive Adaptive Reweighted Sampling [CARS]) on model performance. Through comprehensive comparison of multiple ML classifiers—including Support Vector Machine (SVM), Random Forest (RF), k‐Nearest Neighbors (KNN), and Partial Least Squares Discriminant Analysis (PLS‐DA)—combined with soft voting ensemble learning, we developed an optimized FD‐CARS‐Voting model that achieved exceptional classification accuracy (99.06% ACC, 98.59% Kappa). This approach provides an efficient, non‐destructive solution for real‐time quality monitoring in commercial supply chains, effectively balancing detection accuracy, processing speed, and cost‐effectiveness. The proposed framework demonstrates strong potential for extension to other perishable agricultural products requiring rapid quality assessment.