Prediction of Mild Moldy-Core Disease in Apples Based on Fusion Features of Near-Infrared Transmission Spectroscopy...

In response to the low accuracy of single-method detection of mild moldy-core in apples, a heterogeneous information fusion method based on near-infrared transmission spectroscopy and acoustic vibration technology was proposed. For near-infrared spectral signals, the impact of different preprocessing and feature extraction methods on the modeling performance was analyzed, and the optimal spectral feature bands were determined. For acoustic vibration signals, seven time-domain features and six frequency-domain features were selected using one-way ANOVA. Subsequently, the spectral feature bands and acoustic vibration features were concatenated to form a fused feature vector. Support vector machines (SVM) were employed to construct discrimination models based on single-source features and fused features, and different optimization algorithms were applied to optimize the parameters 'c' and 'g.' Model performance analysis results showed that the FF-SSA-SVM model, which integrates 18 spectral feature bands from near-infrared transmission spectroscopy and 13 time-domain and frequency-domain features, achieved the best performance in discriminating mild moldy-core. Specifically, the accuracy and F1_score reached 97.97% and 97.45% respectively. It demonstrates that the proposed method of fused near-infrared transmission spectroscopy and acoustic vibration features can significantly improve the accuracy of detecting mild moldy-core in apples. Additionally, the AUC of the test set reached 0.9757, indicating that the model has good generalization and robustness for imbalanced datasets. Therefore, this study provides an effective nondestructive testing method for mild moldy-core of apples and can give some research ideas for internal quality detection of other fruits and vegetables. PRACTICAL APPLICATION: This study provides an approach for the precise detection of mild moldy-core in apples during industrial production.