Research on wheat impurity identification method based on terahertz imaging technology.

Area of Science:

• Agricultural Science
• Imaging Technology
• Artificial Intelligence

Background:

• Traditional wheat impurity detection methods are imprecise, slow, and labor-intensive.
• Accurate wheat quality assessment is crucial and relies on effective impurity detection.
• Terahertz (THz) technology offers rapid, non-destructive, and penetrating imaging for material analysis.

Purpose of the Study:

• To develop a rapid and accurate method for detecting and classifying impurities in wheat using THz imaging.
• To introduce and evaluate the AHA-RetinaNet-X algorithm for wheat and impurity identification.
• To enhance wheat quality grading through non-contact, non-destructive analysis.

Main Methods:

• Utilized a THz three-dimensional tomography imaging system to capture images of wheat and impurities.
• Developed a novel classification and recognition algorithm, AHA-RetinaNet-X, integrating RetinaNet with the Artificial Hummingbird Algorithm (AHA).
• Trained and validated the model on two distinct THz image datasets: one for wheat/impurity classification and another for impurity classification.

Main Results:

• The AHA-RetinaNet-X model demonstrated superior performance over other models in accuracy, F1-score, precision, recall, and specificity.
• Achieved high accuracy rates (e.g., 96.1% for wheat/impurity dataset, 95.6% for impurity dataset).
• The model reached a mean Average Precision (mAP) of 92.1%, outperforming comparative models.

Conclusions:

• The integration of THz imaging and the AHA-RetinaNet-X algorithm provides an effective new approach for non-contact, rapid, and non-destructive detection of wheat impurities.
• This method significantly improves the accuracy and efficiency of wheat quality assessment.
• The findings offer a valuable reference for developing similar detection and identification methods for other substances.