Enhancing grain drying methods with hyperspectral imaging technology: A visualanalysis.

Area of Science:

• Agricultural Engineering
• Food Science
• Spectroscopy

Background:

• Grain drying is crucial for storage and quality preservation.
• Different drying methods impact grain characteristics.
• Non-destructive methods for assessing drying effects are needed.

Purpose of the Study:

• To develop and compare models for recognizing different grain drying methods.
• To utilize hyperspectral imaging technology (HSI) and multivariate analysis for this purpose.
• To evaluate the effectiveness of spectral and texture feature fusion.

Main Methods:

• Collected hyperspectral images (388-1065 nm) of grain dried by rotating ventilation, mechanical, and natural methods.
• Extracted spectral features using Principal Component Analysis (PCA).
• Extracted texture features using second-order probability statistical filtering.
• Developed Partial Least Squares Regression (PLSR) and Back-propagation Neural Network (BPNN) models.
• Fused spectral and texture features for enhanced recognition.

Main Results:

• Texture analysis revealed distinct characteristics for mechanical drying compared to other methods.
• The BPNN model using fused spectral-texture features demonstrated superior performance in distinguishing drying modes.
• Pseudo-color visualization effectively represented differences in drying methods based on fused features.

Conclusions:

• Hyperspectral imaging combined with spectral-texture feature fusion provides an effective non-destructive method for identifying grain drying techniques.
• The BPNN model offers a robust approach for rapid and accurate grain quality assessment post-drying.
• This research offers valuable insights for optimizing grain drying processes and quality control.