Study on Lightweight Bridge Crack Detection Algorithm Based on YOLO11

Area of Science:

• Computer Vision
• Artificial Intelligence
• Structural Health Monitoring

Background:

• Traditional bridge crack detection methods lack efficiency and accuracy.
• Deep learning-based computer vision is emerging as a hotspot for bridge crack detection.
• Ensuring bridge safety and lifespan necessitates effective crack detection.

Purpose of the Study:

• To propose a lightweight and efficient bridge crack detection algorithm.
• To enhance feature extraction and detection accuracy without increasing model size.
• To achieve real-time detection and high computational efficiency for bridge monitoring.

Main Methods:

• Optimization of the YOLO11 model to create YOLO11-Bridge Detection (YOLO11-BD).
• Integration of an efficient multiscale conv all (EMSCA) module for enhanced channel and spatial attention.
• Introduction of a lightweight detection head (LDH) utilizing grouped convolutions for reduced parameters and computation.

Main Results:

• YOLO11-BD improved mAP50 by 3.1% and mAP50-95 by 4.8% compared to the original YOLO11.
• GFLOPs were reduced by 19.05%, indicating significant computational efficiency.
• Achieved a detection speed exceeding 500 frames per second, demonstrating excellent real-time capability.

Conclusions:

• The YOLO11-BD algorithm offers an efficient and flexible solution for bridge crack monitoring using remote sensing.
• Its lightweight design ensures cross-platform adaptability for intelligent bridge management.
• Provides reliable technical support for enhancing bridge safety and maintenance through advanced computer vision.