Combination of Spectral and Spatial Information of Hyperspectral Imaging for the Prediction of the Moisture Content and Visualizing Distribution in Daqu

The moisture content and distribution of Daqu significantly influences the quality of Daqu products. This work presents the visualization of the moisture content in Daqu using a combination of spectral and spatial information from hyperspectral imaging. The least squares support vector machine (LS-SVM) and partial least squares regression (PLSR) methods were adopted to establish the predictive models based on the full wavelengths and the 29 feature wavelengths combined with color features, respectively. The best prediction model was PLSR (Rp2=0.9823, RMSEP=0.0109) based on feature wavelengths. The results showed that the combination of spectral and spatial information of hyperspectral imaging can accurately predict the moisture content in Daqu during different fermentation processes, and the visualization of the distribution map of moisture content in Daqu provided a more convenient and understandable assessment of moisture content. This work presents a novel, rapid, and nondestructive approach for moisture content detection in Daqu, and provides theoretical support and basis for intelligent adjustment of temperature, humidity and other environmental parameters of Daqu fermentation.