Hyperspectral imaging and explainable deep-learning for non-destructive quality prediction of sweetpotato

Abstract

The recent combination of hyperspectral imaging (HSI) and deep learning, particularly through convolutional neural networks (CNN), offers a non-destructive and innovative approach for accurate predictive analysis of post-harvest food products. This study utilized HSI and CNN-based regression to predict the firmness of various sweetpotato varieties by extracting spectral data from images captured with a visible near-infrared HSI system (400–1000 nm). The hyperparameters of CNN were fine-tuned using Bayesian Optimization (BO), which resulted in an 18.42 % reduction in the prediction RMSE compared to the traditional partial least squares regression (PLSR) model. Additionally, explainable artificial intelligence techniques were applied to interpret the CNN model and assess the contribution of the variable wavelengths. The CNN model based on important wavelengths was used to visualize the spatial distribution of firmness in sweetpotato samples. The analytical outcomes highlight the efficiency and growing importance of applying HSI with explainable deep learning for advanced post-harvest analyses.