An Efficient Lightweight Object Detection Algorithm for Defect Identification in Paeoniae Radix Alba With Rapid Area...

Area of Science:

• Traditional Chinese Medicine
• Agricultural Science
• Computer Vision

Background:

• Paeoniae Radix Alba (white peony root, WP) is a vital herb in traditional Chinese medicine, food, and health supplements.
• Ensuring the quality of WP is critical for its efficacy, safety, and effectiveness in various applications.
• Efficient and accurate inspection methods are essential for quality control of WP.

Purpose of the Study:

• To develop an efficient algorithm for defect detection and physical dimension measurement of Paeoniae Radix Alba decoction pieces.
• To improve upon existing models like YOLOv8 in terms of accuracy and computational efficiency.
• To ensure the quality and safety of WP for its dual use in food and medicine.

Main Methods:

• A novel detection algorithm incorporating a self-developed LSD-Head and Feature Fusion Attention (FCA) mechanism was employed for defect detection.
• OpenCV technology was utilized for precise measurement of physical dimensions (area, maximum/minimum diameters) of WP slices.
• The proposed model's performance was benchmarked against the YOLOv8 model.

Main Results:

• The proposed algorithm demonstrated superior defect detection capabilities compared to YOLOv8.
• The model achieved a 2.6% accuracy improvement over YOLOv8 while reducing parameter size by 13% and computational load to 65.8% of YOLOv8's.
• Physical dimension measurements showed an average error of less than 5% compared to actual sizes.

Conclusions:

• The developed algorithm effectively detects defects and accurately measures physical dimensions of Paeoniae Radix Alba decoction pieces.
• This method offers a more efficient and accurate quality inspection solution for WP.
• The findings support the reliable application of WP in food products and health supplements.