Related Experiment Video
Updated: Aug 8, 2026
05:30
Crack Monitoring in Resonance Fatigue Testing of Welded Specimens Using Digital Image Correlation
Published on: September 29, 2019
An RPP-YOLOv11 model for road crack detection
Yuhong Xue1, Ligang Zheng1, Yangyang Shi2
1Shanxi Steel Structure Science & Industry Co., Ltd., Taiyuan, Shanxi, China.
Plos One
|August 6, 2026
Summary
This study introduces RPP-YOLOv11, an improved road crack detection model using fused visible and thermal images, specialized convolutions, and an expanded detection framework. It significantly enhances accuracy and efficiency for smart transportation infrastructure maintenance.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Civil Engineering
Background:
- Road crack detection is vital for smart transportation and infrastructure maintenance.
- Existing YOLO models struggle with subtle crack details, variations, and background noise.
- Limitations hinder accurate and efficient automated road inspection.
Purpose of the Study:
- To enhance road crack detection accuracy and efficiency using an improved YOLOv11 model.
- To address limitations in existing models regarding subtle details, morphological variations, and background interference.
- To provide a robust technical solution for automated road inspection systems.
Main Methods:
- Developed RPP-YOLOv11 by integrating RGBT (visible and thermal infrared) multispectral fusion.
- Implemented windmill-shaped convolution modules (PSConv) for superior feature extraction and noise suppression.
- Introduced a P6 detection layer, creating a four-scale detection framework (P3-P6) for enhanced perception.
Main Results:
- RPP-YOLOv11 achieved 74.90% accuracy, 64.36% recall, and 69.04% mAP@0.5 on the RDD2022 dataset.
- Demonstrated significant improvements over original YOLOv11 and other benchmarks in precision and robustness.
- Showcased enhanced detection of cracks with diverse morphologies and complex backgrounds.
Conclusions:
- The proposed RPP-YOLOv11 model offers a reliable and efficient solution for automated road inspection.
- RGBT fusion, PSConv, and the P6 detection layer collectively boost detection performance.
- This advancement contributes to smarter transportation systems and proactive infrastructure maintenance.
