Soluble solid content (SSC) is a critical index to evaluate the nutrition and flavor quality of food products. This study presents a novel strategy to predict the SSC in Agaricus bisporus slices during ultrasound-assisted osmotic dehydration (UOD). The spectral signatures of Agaricus bisporus were captured via a hyperspectral imaging (HSI) system and different spectral preprocessing methods and models were used to fit and evaluate the SSC behaviour of samples during UOD. The results showed that the support vector machine (SVM) preprocessed with orthogonal signal correction (OSC) provided the best fit for the full-band spectra of samples, with a higher correlation coefficient of prediction (R2 P, 0.883) and residual predictive deviation (RPD, 3.04). Moreover, the competitive adaptive reweighted sampling (CARS) algorithm can screen 67 key wavelengths from the complex original full-band wavelengths, and the OSC-CARS-SVM model showed the best predicted performance of SSC for the simplified spectra. In addition, the distribution of SSC in different UOD periods of the samples were demonstrated in a pseudo-colour map, which further revealed the SSC distribution of samples during UOD. The overall results showed the great potential of HSI to detect and predict the SSC of Agaricus bisporus rapidly, accurately, and non-destructively.

可溶性固形物（SSC）是评估食品营养和风味品质的关键指标。这项研究提出了一种新的策略，可以预测双孢蘑菇切片在超声辅助渗透脱水（UOD）过程中的SSC 。通过高光谱成像（HSI）系统捕获双孢蘑菇的光谱特征，并使用不同的光谱预处理方法和模型拟合和评估UOD期间样品的SSC行为。结果表明，经过正交信号校正（OSC）预处理的支持向量机（SVM）最适合样本的全波段光谱，并且具有更高的预测相关系数（R2 P，0.883）和剩余预测偏差（RPD，3.04）。此外，竞争性自适应加权采样（CARS）算法可以从复杂的原始全波段波长中筛选出67个关键波长，并且OSC-CARS-SVM模型对于简化光谱显示出SSC的最佳预测性能。另外，在伪彩色图中证明了样品在不同的UOD期间SSC的分布，这进一步揭示了UOD期间样品的SSC分布。总体结果表明，HSI能够快速，准确，无损地检测和预测双孢蘑菇的SSC 。