A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Road Pavement Crack Detection Algorithm Based on Vehicle Panoramic Image
Pavement cracks are an important indicator of highway maintenance. Timely and accurate detection of pavement cracks is crucial for formulating effective maintenance plans and ensuring road safety. However, due to the irregular sizes and random shapes of cracks, automatic detection of pavement cracks remains a challenging task. In this paper, a new crack detection method based on deep learning is introduced for urban pavement crack detection. Firstly, a new crack labeling criterion is proposed, and experiments show that the best detection performance of the model is achieved when the crack width is within 5 times the average crack width from the center point of the marked frame. The width-height ratio of the marked frame does not exceed 4. Then, the regression stage of the YOLOv5 network is improved using an anchor-free method to address the issue of model perform ance degradation caused by unreasonable selection of YOLOv5 network anchors. The results show that compared with existing mainstream solutions, the proposed method has significant advantages. The accuracy of the new model is more than 81%, while the accuracy of the baseline method is less than 76%.
Road Pavement Crack Detection Algorithm Based on Vehicle Panoramic Image
Pavement cracks are an important indicator of highway maintenance. Timely and accurate detection of pavement cracks is crucial for formulating effective maintenance plans and ensuring road safety. However, due to the irregular sizes and random shapes of cracks, automatic detection of pavement cracks remains a challenging task. In this paper, a new crack detection method based on deep learning is introduced for urban pavement crack detection. Firstly, a new crack labeling criterion is proposed, and experiments show that the best detection performance of the model is achieved when the crack width is within 5 times the average crack width from the center point of the marked frame. The width-height ratio of the marked frame does not exceed 4. Then, the regression stage of the YOLOv5 network is improved using an anchor-free method to address the issue of model perform ance degradation caused by unreasonable selection of YOLOv5 network anchors. The results show that compared with existing mainstream solutions, the proposed method has significant advantages. The accuracy of the new model is more than 81%, while the accuracy of the baseline method is less than 76%.
Road Pavement Crack Detection Algorithm Based on Vehicle Panoramic Image
Liu, Rufei (author) / Lai, Ruixin (author) / Ren, Hongwei (author) / Zhang, Yi (author) / Li, Zeyu (author) / Su, Zhanwen (author)
2024-09-20
18919176 byte
Conference paper
Electronic Resource
English
Automatic road crack detection from pavement images
| DataCite | 2024
Automatic pavement crack detection based on image recognition
| TIBKAT | 2019
Crack detection device for road and bridge asphalt pavement
| European Patent Office | 2022