In this study, visible-near infrared (Vis-NIR) spectra of Xinjiang plums at four different shelf-life stages (1, 3, 5, and 7 d after harvest) were collected. Four spectral preprocessing methods, Savitzky-Golay (SG) smoothing, standard normal variate transformation (SNV), baseline correction (BC), and min-max normalization (MN), were compared to develop shelf-life classification models using partial least squares-discriminant analysis (PLS-DA) or extreme learning machine (ELM). For the development of PLS-DA and ELM models, feature wavelength extraction methods, namely competitive adaptive reweighted sampling (CARS), variable iterative space shrinkage approach (VISSA), and successive projection algorithm (SPA), were separately combined with three spectral features, namely band ratio (BR), band difference (BD), and normalized spectral intensity difference (NSID), thereby forming combined feature sets. The results showed that the PLS-DA and ELM models based on the full spectrum exhibited limited classification performance with a validation accuracy of only 70.24%. In contrast, the MN-SPA-NSID-ELM model, which integrated MN preprocessing, SPA-based feature extraction, and NSID, achieved a validation accuracy of 97.62% using only 24 selected variables and 30 hidden layer neurons, significantly outperforming the other combined models. In addition, on an independent external test set of Xinjiang plums, this model still achieved an accuracy of 99.18% with a Kappa coefficient of 0.989, indicating improved shelf-life classification efficiency. This study provides a rapid, accurate, and nondestructive technique for the production and grading of Xinjiang plums.

In response to the issue of low detection accuracy for mild moldy-core disease in apples using single methods, this study proposed an approach based on the fusion of near-infrared transmission spectroscopy and acoustic-vibration technology to enhance the discriminability of mild moldy-core disease in apples. For the near-infrared spectral signals, the impacts of different preprocessing and feature extraction methods on modeling outcomes were analyzed to select the spectral feature bands. For the acoustic-vibration signals, 7 time-domain features were optimized by using the YSV engineering test and signal analysis software as well as calculating Pearson correlation coefficients. The spectral feature bands and time-domain features were then concatenated to form a fused feature vector. Convolutional neural networks (CNN), long short-term memory (LSTM), and CNN-LSTM were employed to construct discrimination models based on single and fused features, separately. The performance analysis of the models revealed that the CNN-LSTM combination model, which integrated 15 near-infrared transmission spectral bands and 7 time-domain features, exhibited the best performance in discriminating mild moldy-core disease, with accuracy, recall, specificity, and F1 scores of 98.31%, 97.06%, 97.06%, and 97.90% on the test set, respectively. These findings demonstrate that the proposed method can effectively improve the discrimination accuracy of mild moldy-core disease in apples.

To address the difficulty of detecting moldy walnuts and the problem of low detection efficiency, a non-destructive method based on fused features of X-ray and visual images was proposed to accurately distinguish four grades of moldy walnuts: moldy both internally and externally, moldy internally and normal externally, normal internally and moldy externally, and normal both internally and externally. First, the gray-level co-occurrence matrix (GLCM) was used to extract texture features from X-ray and visual images, and the first and second moments of the visual images were computed in different color spaces to comprehensively capture the internal and external moldiness characteristics of walnuts in order to construct an original moldy walnut feature set. Subsequently, using competitive adaptive reweighted sampling (CARS) and successive projection algorithm (SPA), the extracted features were optimized to construct a walnut feature set sensitive to different degrees of moldiness. On this basis, an extreme learning machine (ELM) model and a K-nearest neighbors (KNN) model were developed for moldy walnut classification, and the performance of the classification models under different feature sets was compared through experiments to verify the feasibility of fusing X-ray and visual image features for detecting moldy walnuts. The experimental results showed that the ELM model developed using SPA optimized feature set had the best performance. The accuracy and recall for the test set, and the harmonic mean of precision and recall (F₁) value of the model were 90.32%, 92.58%, and 91.29%, respectively. The average specificity and Kappa coefficient values were 97.02% and 88.44%, respectively, indicating high ability to discriminate both majority and minority moldy walnuts. This study provides a theoretical reference for the comprehensive and accurate identification of the internal and external moldiness of walnuts, as well as the development of online non-destructive testing systems.