Advanced Detection of Black Tea Fermentation Quality Using Hyperspectral Imaging and Optimized Backpropagation Neural Network

This study investigates the application of hyper- spectral imaging combined with an optimized Backpropagation Neural Network (BPNN) for evaluating black tea fermentation quality. Spectral data from tea leaves at five fermentation stages were acquired and analyzed using feature extraction methods such as Successive Projections Algorithm (SPA), Variable Iterative Space Shrinkage Approach (VISSA), and Competitive Adaptive Reweighted Sampling (CARS). Among various optimization algorithms, the Whale Optimization Algorithm (WOA) - BPNN model achieved the highest prediction accuracy of 92.0% under five-fold cross-validation. The VISSA-WOA-BPNN model demonstrated efficient, nondestructive, and online detection capabilities. This approach provides a robust and accurate method for assessing black tea fermentation quality.