改进YOLOv5s的路面坑槽目标检测模型

doi:10.16265/j.cnki.issn1003-3033.2025.01.0619

摘要/Abstract

摘要：

为提高道路安全巡检工作中路面坑槽隐患的检测效率和自动化水平,降低交通事故发生概率,构建一种基于改进YOLOv5s的路面坑槽隐患智能检测模型。在原YOLOv5s网络中加入自适应空间特征融合(ASFF)模块,将主干网络替换为FasterNet网络,引入轻量通道注意力(ECA)模块;通过消融试验分析改进模块对检测模型性能的影响,验证目标检测效果,并开发交互式可视化检测界面。结果表明: 改进后的模型精度、召回率和平均检测精度分别提升了4.1%、9.9%和5.6%。较原网络有较为显著的提升,具有良好的检测效果,能够满足路面坑槽自动化检测的应用需求,提高巡检效率,减少因路面坑槽导致的交通事故。

关键词: YOLOv5s, 路面坑槽, 目标检测, 自适应空间特征融合(ASFF), FasterNet

Abstract:

To improve the detection efficiency and automation level of detecting road surface pits and grooves in road safety inspection work, reduce the probability of traffic accidents. A road surface pit and groove hazard intelligent detection model based on an improved YOLOv5s was proposed. This method incorporated the ASFF module into the original YOLOv5s network, replaced the backbone network with the FasterNet network, and further introduced the Efficient Channel Attention (ECA) module. Ablation experiments are conducted to analyze the effect of the improved module on performance of the detection model, to verify the target detection effect, and to develop an interactive visualized detection interface. The results show that the improved model accuracy, recall rate, and average detection accuracy have increased by 4.1%, 9.9% and 5.6% respectively. Compared to the original network, the improvement is significant. It demonstrats good detection performance that meets the application requirements for automated detection of road surface pits and grooves, thereby enhancing inspection efficiency and effectively reducing traffic accidents caused by road surface pits and grooves.

Key words: YOLOv5s, pavement potholes, target detection, adaptive spatial feature fusion(ASFF), FasterNet

中图分类号:

X924.3安全监控系统

赵江平, 王欣然, 吴立舟. 改进YOLOv5s的路面坑槽目标检测模型[J]. 中国安全科学学报, 2025, 35(1): 67-74.

ZHAO Jiangping, WANG Xinran, WU Lizhou. Model of pavement pothole target detection with improved YOLOv5s[J]. China Safety Science Journal, 2025, 35(1): 67-74.

图/表 10

图1

图2

图3

图4

图5

表1

图6

图7

图8

图9

参考文献 15

[1]	道路交通管理杂志社. 数说交通 \| 2023年, 全国公路总里程544.1万公里, 其中高速公路18.4万公里[EB/OL]. (2024-04-08). https://www.yoojia.com/article/9539097732664945371.html.
[2]	纪雅琪, 程博文, 龚煜. 智能道路状态检测系统研究[J]. 长江信息通信, 2023, 36(3):154-156,159.
	JI Yaqi, CHENG Bowen, GONG Yu. Research on intelligent road state detection system[J]. Changjiang Information & Communications, 2023, 36(3):154-156,159.
[3]	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016: 779-788.
[4]	高倩. 公路路面病害智能检测技术研究与应用[D]. 北京: 北京工业大学, 2022.
	GAO Qian. Research and application of intelligent detection technology for highway pavement disease[D]. Beijing: Beijing University of Technology, 2022.
[5]	白芮, 徐杨, 王彬, 等. 基于改进YOLOv5s的道路坑洼检测算法[J]. 计算机与现代化, 2023(6):69-75.
	BAI Rui, XU Yang, WANG Bin, et al. Road pothole detection algorithm based on improved YOLOv5s[J]. Computer and Modernization, 2023(6):69-75.
[6]	朱瑞鑫, 杨福兴. 运动场景下改进YOLOv5小目标检测算法[J]. 计算机工程与应用, 2023, 59(10):196-203.
	ZHU Ruixin, YANG Fuxing. Improved YOLOv5 small object detection algorithm in moving scenes[J]. Computer Engineering and Applications, 2023, 59(10):196-203.
[7]	钟靖涛. 基于无人机图像与深度学习的路面病害识别研究[D]. 南京: 东南大学, 2022.
	ZHONG Jingtao. Research on pavement disease recognition based on UAV images and deep learning[D]. Nanjing: Southeast University, 2022.
[8]	高明星, 关雪峰, 范井丽, 等. 基于改进YOLOv5-DeepSORT算法的公路路面病害智能识别[J]. 森林工程, 2023, 39(5):161-174.
	GAO Mingxing, GUAN Xuefeng, FAN Jingli, et al. Intelligent recognition of road surface disease based on improved YOLOv5-DeepSORT algorithm[J]. Forest Engineering, 2023, 39(5):161-174.
[9]	李雪. 公路路面坑槽及应对策略对交通的影响[D]. 北京: 北京交通大学, 2022.
	LI Xue. Impact of highway pavement potholes and coping strategies on traffic[D]. Beijing: Beijing Jiaotong University, 2022.
[10]	LIU Songtao, HUANG Di, WANG Yunhong. Learning spatial fusion for single-shot object detection[J]. ArXiv Preprint ArXiv, 2019: DOI: 10.48550/arXiv.1911.09516.
[11]	CHEN Jierun, KAO Shiuhong, HE Hao, et al. Run, don't walk: chasing higher FLOPS for faster neural networks[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023:12 021-12 031.
[12]	WANG Qilong, WU Banggu, ZHU Pengfei, et al. ECA-Net: efficient channel attention for deep convolutional neural networks[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020:11 531-11 539.
[13]	刘雅洁, 伊力哈木·亚尔买买提, 席凌飞, 等. 改进YOLOv5s的安全帽佩戴检测算法研究[J]. 计算机工程与应用, 2023, 59(20):184-191.
	LIU Yajie, YILHAMU Yaermaimaiti, XI Lingfei, et al. Research on improved safety helmet wearing detection algorithm of YOLOv5s[J]. Computer Engineering and Applications, 2023, 59(20):184-191.
[14]	赵江平, 刘星星, 张想卓. 基于改进YOLOv5s的外脚手架隐患图像识别技术[J]. 中国安全科学学报, 2023, 33(12):60-66.
	ZHAO Jiangping, LIU Xingxing, ZHANG Xiangzhuo. Research on image recognition technology for external scaffold hidden danger based on improved YOLOv5s[J]. China Safety Science Journal, 2023, 33(12):60-66.
[15]	李华, 吴立舟, 薛曦澄, 等. 基于计算机视觉的高处临边作业安全巡检[J]. 中国安全科学学报, 2023, 33(9):69-75.
	LI Hua, WU Lizhou, XUE Xicheng, et al. Computer vision based safety inspection of high abutting edges[J]. China Safety Science Journal, 2023, 33(9):69-75.

模型	ASFF	FasterNet	ECA	P/%	R/%	mAP@ 0.5/%	mAP@0.5∶ 0.95/%	FPS/ (帧·s^-1)	浮点运算量/G	参数量/ MiB
YOLOv5s	—	—	—	85.0	70.7	80.9	52.7	176.8	15.8	7.01
YOLOv5s-A	√	—	—	87.3	72.6	81.6	54.6	149.6	24.2	12.45
YOLOv5s-F	—	√	—	83.8	76.2	82.3	53.3	179.9	13.7	5.99
YOLOv5s-E	—	—	√	83.1	77.6	82.6	54.8	175.5	15.8	7.01
YOLOv5s-AF	√	√	—	86.8	76.3	83.1	57.2	154.8	24.2	12.45
YOLOv5s-AFE	√	√	√	89.1	80.6	86.5	59.5	146.0	22.1	11.42