Scholarlink

Aiming at the problem of urban road disease detection, this paper proposes a method based on the combination of improved YOLOv11 detection and threshold crack segmentation, which can effectively improve the recognition and extraction of small road diseases. The main contribution of this paper is to introduce the ECA attention mechanism of crack direction into the backbone network and use the improved YOLOv11 algorithm to detect images, which effectively improves the accuracy of crack detection. The multi-threshold method is used to segment and extract cracks. The experimental results show that the average precision, recall rate and mAP @50 value of the yolov11 + ECA network model are significantly improved; compared with the segmentation results of the full threshold algorithm and the adaptive threshold algorithm, the multi-threshold algorithm has complete and clear cracks. The combination of the above two methods makes the road crack detection more effective, with a high degree of automated detection and accurate crack extraction. This test method helps to improve the intelligence of road crack detection and has better prospects in the future.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Crack Detection Technology for Urban Asphalt Roads Based on an Improved YOLOv11 Network Model with Increased ECA Attention Mechanism

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Zhong Wang,
  • Li Fan,
  • Jiancheng Zou

摘要

Aiming at the problem of urban road disease detection, this paper proposes a method based on the combination of improved YOLOv11 detection and threshold crack segmentation, which can effectively improve the recognition and extraction of small road diseases. The main contribution of this paper is to introduce the ECA attention mechanism of crack direction into the backbone network and use the improved YOLOv11 algorithm to detect images, which effectively improves the accuracy of crack detection. The multi-threshold method is used to segment and extract cracks. The experimental results show that the average precision, recall rate and mAP @50 value of the yolov11 + ECA network model are significantly improved; compared with the segmentation results of the full threshold algorithm and the adaptive threshold algorithm, the multi-threshold algorithm has complete and clear cracks. The combination of the above two methods makes the road crack detection more effective, with a high degree of automated detection and accurate crack extraction. This test method helps to improve the intelligence of road crack detection and has better prospects in the future.