Research on the crack detection method of black coating based on machine vision and deep learning

Area of Science:

• Materials Science
• Computer Vision
• Artificial Intelligence

Background:

• Thermal stress in black high-radiation coatings on porous materials can cause micro-cracks, compromising structural integrity.
• Low-contrast imaging of these small cracks hinders accurate real-time detection.

Purpose of the Study:

• To develop a machine vision system for improved detection of small cracks in black coatings.
• To enhance the accuracy and efficiency of crack detection algorithms for structural integrity monitoring.

Main Methods:

• Investigated lighting effects on crack-background contrast.
• Developed a crack detection dataset using data augmentation and annotation.
• Introduced the BCC-YOLO algorithm, modifying YOLOv10s with an ADown module and iEMA attention mechanism.
• Utilized the UIoU loss function for improved training stability and convergence.

Main Results:

• BCC-YOLO achieved significant improvements in precision (9.7%), recall (11.2%), mAP50 (10.8%), and mAP50:95 (9.8%) compared to YOLOv10s.
• Reduced computational complexity by 7.3% (FLOPs).
• Demonstrated enhanced feature extraction for small cracks and improved detection accuracy.

Conclusions:

• The proposed machine vision system and BCC-YOLO algorithm effectively address low-contrast crack detection challenges in black coatings.
• The study offers a high-precision, computationally efficient solution for automatic crack detection, crucial for structural safety applications.