Related Experiment Video
Updated: Apr 15, 2026
11:34
Subsurface Defect Localization by Structured Heating Using Laser Projected Photothermal Thermography
Published on: May 15, 2017
11.6K
Digital Twin-Based Improved YOLOv8 Algorithm for Micro-Defect Detection of Labyrinth Drip Emitters in High-Speed
Renzhong Niu1, Zhangliang Wei2, Peilin Jin2
1College of Mechanical and Electrical Engineering, Shihezi University, Shihezi 832000, China.
Sensors (Basel, Switzerland)
|April 14, 2026
Summary
This study introduces an advanced defect detection system for drip irrigation emitters, integrating Digital Twin technology and an enhanced YOLOv8 model. The system ensures reliable water-saving agriculture by improving irrigation uniformity and water-use efficiency in arid regions.
Area of Science:
- Agricultural Engineering
- Computer Vision
- Artificial Intelligence
Background:
- Water scarcity in regions like Xinjiang necessitates advanced water-saving irrigation technologies.
- Defects in labyrinth drip emitters, crucial for drip irrigation, compromise system reliability and irrigation uniformity.
- Existing detection methods may not adequately address micro-defects in manufacturing.
Purpose of the Study:
- To develop an online defect detection approach for labyrinth drip emitters using Digital Twin (DT) technology and an enhanced YOLOv8 model.
- To improve the accuracy and efficiency of identifying micro-defects in drip irrigation tape production lines.
- To enhance water-use efficiency and support sustainable agriculture in arid regions.
Main Methods:
- Integration of Digital Twin (DT) framework with a refined YOLOv8 model for online inspection.
- Development of a custom dataset with six defect categories for training and validation.
- Implementation of DySnakeConv, DySample, Efficient Multi-Scale Attention, and Inner-SIoU for enhanced defect detection capabilities.
Main Results:
- The proposed model achieved 89.6% precision, 90.9% recall, and 93.9% mAP50, outperforming baseline YOLOv8, YOLOv10, and YOLOv11.
- Significant improvements in precision (up to 10.0%), recall (up to 7.7%), and mAP50 (up to 7.3%) compared to baseline YOLOv8.
- The model maintains a lightweight structure (3.7M parameters) and real-time inference speed (150.2 FPS), indicating a strong accuracy-efficiency trade-off.
Conclusions:
- The DT-integrated enhanced YOLOv8 model provides a robust solution for online defect detection of labyrinth drip emitters.
- The approach effectively addresses micro-defects, ensuring higher quality components for drip irrigation systems.
- This technology offers practical support for precision agriculture by enhancing irrigation uniformity and water-use efficiency in water-scarce environments.
