Quality grading method for Pleurotus eryngii during postharvest storage based on hyperspectral imaging and multiple quality indicators

During storage, Pleurotus eryngii is prone to spoilage, significantly affecting their commercial value. This study addresses the issue of inconsistent quality grading standards and low detection accuracy for Pleurotus eryngii. . It proposes a method for grading the quality of Pleurotus eryngii during post-harvest low-temperature (4 degrees C) storage, achieving rapid quality grading based on hyperspectral imaging technology combined with clustering algorithms and comprehensive evaluation methods. According to 11 physical and chemical indexes analyzed by K-means algorithm and principal component analysis, the samples were divided into three categories: high, medium and inferior. These clustering results were then used as labels to build classification models. We compared the feature extraction capabilities of convolutional autoencoder and competitive adaptive reweighted sampling, as well as the modeling effectiveness of support vector machine, linear discriminant analysis, and partial least squares discriminant analysis. The models' practicality was validated using external data. The results showed that the convolutional autoencoder-support vector machine model performed the best in quality classification, with an accuracy of 91.58%, an F1 score of 91.36%, a precision of 89.65%, and a recall of 90.60%. In summary, this study demonstrates the rationality of using clustering algorithms for quality grading of Pleurotus eryngii and highlights the significant improvement in prediction accuracy and generalization ability achieved by convolutional autoencoder.