Research on Road Defect Detection Algorithm Based on YOLOv11 with Multi-scale Feature Extraction.

Saved in:

Bibliographic Details
Title:	Research on Road Defect Detection Algorithm Based on YOLOv11 with Multi-scale Feature Extraction.
Authors:	Liu, Yong¹ 1762431278@qq.com, Tian, Ying² astianying@126.com
Source:	Engineering Letters. Jul2026, Vol. 34 Issue 7, p2877-2887. 11p.
Subjects:	Feature extraction, Object recognition (Computer vision), Real-time computing
Abstract:	To address the difficulties in feature extraction and the low detection accuracy caused by irregularly shaped and variably sized road defects such as cracks and potholes, an improved multi-scale feature extraction algorithm for road defect detection, termed YOLOv11m-CRC, is proposed based on YOLOv11m. First, the YOLOv11m-CRC algorithm designs a novel C3k2_SDSConv module, in which dynamic snake convolution (DSConv) and dilated convolution (SConv) are introduced into the C3k2 module, thereby enlarging the receptive field without increasing the number of parameters and enhancing the model's ability to extract irregular features. Second, a new CRC module is designed by incorporating the lightweight CARAFE operator and residual connections into the Neck to replace the conventional Sample upsampling module, thereby improving the extraction of low-frequency information and increasing network depth. Finally, a deformable bi-level routing attention (DBRA) mechanism is introduced into the C2PSA module to strengthen feature extraction capability and reduce redundant information. Experimental results show that, on the RDD2020 dataset, the improved algorithm achieves a Precision of 60.9%, a mean Average Precision (mAP) of 56.2%, and a Recall of 53.0%, representing improvements of 0.6%, 2.3%, and 2.1%, respectively, over the YOLOv11m algorithm. These results basically satisfy the requirements of real-time performance and robustness in road defect detection, providing an efficient solution for this task. [ABSTRACT FROM AUTHOR]
	Copyright of Engineering Letters is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	To address the difficulties in feature extraction and the low detection accuracy caused by irregularly shaped and variably sized road defects such as cracks and potholes, an improved multi-scale feature extraction algorithm for road defect detection, termed YOLOv11m-CRC, is proposed based on YOLOv11m. First, the YOLOv11m-CRC algorithm designs a novel C3k2_SDSConv module, in which dynamic snake convolution (DSConv) and dilated convolution (SConv) are introduced into the C3k2 module, thereby enlarging the receptive field without increasing the number of parameters and enhancing the model's ability to extract irregular features. Second, a new CRC module is designed by incorporating the lightweight CARAFE operator and residual connections into the Neck to replace the conventional Sample upsampling module, thereby improving the extraction of low-frequency information and increasing network depth. Finally, a deformable bi-level routing attention (DBRA) mechanism is introduced into the C2PSA module to strengthen feature extraction capability and reduce redundant information. Experimental results show that, on the RDD2020 dataset, the improved algorithm achieves a Precision of 60.9%, a mean Average Precision (mAP) of 56.2%, and a Recall of 53.0%, representing improvements of 0.6%, 2.3%, and 2.1%, respectively, over the YOLOv11m algorithm. These results basically satisfy the requirements of real-time performance and robustness in road defect detection, providing an efficient solution for this task. [ABSTRACT FROM AUTHOR]
ISSN:	1816093X