EMF-YOLO：轻量化多尺度特征提取路面缺陷检测算法

doi:10.3778/j.issn.1002-8331.2412-0348

摘要/Abstract

摘要： 道路表面缺陷检测是保障行车安全和延长道路使用寿命的重要技术。现有的道路缺陷检测算法在处理复杂背景、实时性及内存占用方面存在局限性。为此，提出一种基于YOLOv8n的轻量化改进算法EMF-YOLO，旨在提升检测精度并减少计算和内存开销。引入增强型特征融合金字塔EFFPN（enhanced feature fusion pyramid network），优化特征融合路径，提升多尺度特征表示能力。结合可变形注意力机制增强复杂场景下的特征提取能力，并通过多尺度边缘敏感性增强模块MESA（multi-scale edge sensitivity augmentation）替代传统C2f卷积，增强小目标检测能力。设计基于解耦批归一化的共享卷积检测头DBSCD（decoupled bn shared convolution detection head），显著降低模型的参数量和计算复杂度,进一步减小模型体积并加快推理速度。实验结果表明，EMF-YOLO在RDD2022数据集上的检测精度达到了89.2%，较YOLOv5n提高了2个百分点，模型参数量和计算量分别减少了36.1%和25%，在提高检测精度的同时实现较好的轻量化性能。

关键词: 路面缺陷检测, YOLOv8, 轻量化, 多尺度特征提取, 边缘敏感性增强

Abstract: Road surface defect detection is vital for driving safety and extending road lifespan. Existing algorithms struggle with complex backgrounds, real-time processing, and memory usage. This paper presents EMF-YOLO, an improved lightweight algorithm based on YOLOv8n, aiming to enhance detection accuracy while reducing computational and memory costs. The method introduces an enhanced feature fusion pyramid (EFFP) to optimize multi-scale feature representation, and incorporates a deformable attention mechanism (DA) to improve feature extraction in complex scenes. A multi-scale edge sensitivity enhancement (MESA) module replaces traditional C2f convolution to enhance small object detection. Additionally, a decoupled batch normalization-based shared convolution detection head (DBSCD) reduces model parameters and computational complexity, decreasing model size and speeding up inference. Experimental results show that EMF-YOLO achieves 89.2% detection accuracy on the RDD2022 dataset, outperforming YOLOv5n by 2 percentage points, while reducing model parameters and computation by 36.1% and 25%, respectively, balancing accuracy with lightweight performance.

Key words: road surface defect detection, YOLOv8, lightweight, multi-scale feature extraction, edge sensitivity enhancement

秦乐, 谭泽富, 雷国平, 陈秋伯. EMF-YOLO：轻量化多尺度特征提取路面缺陷检测算法[J]. 计算机工程与应用, 2025, 61(14): 101-111.

QIN Le, TAN Zefu, LEI Guoping, CHEN Qiubo. EMF-YOLO: Lightweight Multi-Scale Feature Extraction Algorithm for Road Surface Defect Detection[J]. Computer Engineering and Applications, 2025, 61(14): 101-111.

参考文献

[1] JING J F, LIU S J, WANG G, et al. Recent advances on image edge detection: a comprehensive review[J]. Neurocomputing, 2022, 503: 259-271.
[2] POOTHERI S, ELLAM D, GRüBL T, et al. A two-stage automatic color thresholding technique[J]. Sensors (Basel), 2023, 23(6): 3361.
[3] 王宇瑞, 李炎亮, 郭雨婷, 等. 基于改进霍夫变换的车道线检测算法实现[J]. 农业装备与车辆工程, 2022, 60(5): 95-98.
WANG Y R, LI Y L, GUO Y T, et al. Lane detection algorithm based on improved Hough transform[J]. Agricultural Equipment & Vehicle Engineering, 2022, 60(5): 95-98.
[4] ABU-RADDAHA A, EL-SHAIR Z A, RAWASHDEH S. Leveraging perspective transformation for enhanced pothole detection in autonomous vehicles[J]. Journal of Imaging Science and Technology, 2024, 10(9): 227.
[5] 陈首彬. 激光LiDAR/视觉融合的SLAM (LV-SLAM) 关键技术研究[J]. 测绘学报, 2023, 52(1): 169.
CHEN S B. Research on SLAM based on LiDAR/visual fusion (LV-SLAM)[J]. Acta Geodaetica et Cartographica Sinica, 2023, 52(1): 169.
[6] RABBANI A A M, HUSSAIN M. YOLOv1 to YOLOv10: a comprehensive review of YOLO variants and their application in the agricultural domain[J]. arXiv:2406.10139, 2024.
[7] KUMAR A, ZHANG Z J, LYU H B. Object detection in real time based on improved single shot multi-box detector algorithm[J]. EURASIP Journal on Wireless Communications and Networking, 2020, 2020(1): 204.
[8] WANG Y Y, WANG C, ZHANG H, et al. Automatic ship detection based on RetinaNet using multi-resolution Gaofen-3 imagery[J]. Remote Sensing, 2019, 11(5): 531.
[9] XIE X X, CHENG G, WANG J B, et al. Oriented R-CNN and beyond[J]. International Journal of Computer Vision, 2024, 132(7): 2420-2442.
[10] ZHANG Y, LU Y, HUO Z, et al. USSC-YOLO: enhanced multi-scale road crack object detection algorithm for UAV image[J]. Sensors (Basel), 2024, 24(17): 5586.
[11] WANG X, GAO H, JIA Z, et al. BL-YOLOv8: an improved road defect detection model based on YOLOv8[J]. Sensors (Basel), 2023, 23(20): 8361.
[12] XING Y, HAN X, PAN X, et al. EMG-YOLO: road crack detection algorithm for edge computing devices[J]. Front Neurorobot, 2024, 18: 1423738.
[13] SU P, HAN H Z, LIU M, et al. MOD-YOLO: rethinking the YOLO architecture at the level of feature information and applying it to crack detection[J]. Expert Systems with Applications, 2024, 237: 121346.
[14]王雪秋, 高焕兵, 郏泽萌. 改进YOLOv8的道路缺陷检测算法[J]. 计算机工程与应用, 2024, 60(17): 179-190.
WANG X Q, GAO H B, JIA Z M. Improved road defect detection algorithm based on YOLOv8[J]. Computer Engineering and Applications, 2024, 60(17): 179-190.
[15] 王启涵, 刘超. 改进YOLOv7-Tiny的道路裂缝检测算法[J]. 计算机工程与应用, 2025, 61(10): 372-380.
WANG Q H, LIU C. Improved YOLOv7-Tiny road crack detection algorithm[J]. Computer Engineering and Applications, 2025, 61(10): 372-380.
[16] 胥铁峰, 黄河, 张红民, 等. 基于改进YOLOv8的轻量化道路病害检测方法[J]. 计算机工程与应用, 2024, 60(14): 175-186.
XU T F, HUANG H, ZHANG H M, et al. Lightweight road damage detection method based on improved YOLOv8[J]. Computer Engineering and Applications, 2024, 60(14): 175-186.
[17] 苏俊楷, 段先华, 叶赵兵. 改进YOLOv5算法的玉米病害检测研究[J]. 计算机科学与探索, 2023, 17(4): 933-941.
SU J K, DUAN X H, YE Z B. Research on corn disease detection based on improved YOLOv5 algorithm[J]. Journal of Frontiers of Computer Science and Technology, 2023, 17(4): 933-941.
[18] 马磊, 李晔, 王宇翔. YOLOv8-FD: YOLOv8改进的钢板表面缺陷检测方法[J]. 计算机工程与应用, 2024, 60(24): 211-221.
MA L, LI Y, WANG Y X. YOLOv8-FD: YOLOv8 improved method for detecting surface defects on steel plates[J]. Computer Engineering and Applications, 2024, 60(24): 211-221.
[19] 彭自然, 王思远, 肖伸平. 基于YOLOv8优化改进的太阳能电池片缺陷检测模型[J]. 光电工程, 2024, 51(11): 89-103.
PENG Z R, WANG S Y, XIAO S P. A solar cell defect detection model optimized and improved based on YOLOv8[J]. Opto-Electronic Engineering, 2024, 51(11): 89-103.
[20] XIA Z, PAN X, SONG S, et al. Vision transformer with deformable attention[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022: 4794-4803.
[21] DOHERTY J, GARDINER B, KERR E, et al. BiFPN-YOLO: one-stage object detection integrating bi-directional feature pyramid networks[J]. Pattern Recognition, 2025, 160: 111209.
[22] ARYA D, MAEDA H, GHOSH S K, et al. RDD2022: a multi‐national image dataset for automatic road damage detection[J]. Geoscience Data Journal, 2024, 11(4): 846-862.
[23] YANG L, ZHANG R Y, LI L, et al. SimAM: a simple, parameter-free attention module for convolutional neural networks[C]//Proceedings of the International Conference on Machine Learning, 2021: 11863-11874.
[24] WAN D H, LU R S, SHEN S Y, et al. Mixed local channel attention for object detection[J]. Engineering Applications of Artificial Intelligence, 2023, 123: 106442.
[25] HU S, GAO F, ZHOU X, et al. Hybrid convolutional and attention network for hyperspectral image denoising[J]. IEEE Geoscience and Remote Sensing Letters, 2024, 21: 1-5.
[26] ZHU L, WANG X J, KE Z H, et al. BiFormer: vision transformer with bi-level routing attention[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 10323-10333.
[27] ZHANG Z X, WANG M W. Convolutional neural network with convolutional block attention module for finger vein recognition[J]. arXiv:2202.06673, 2022.
[28] OUYANG D L, HE S, ZHANG G Z, et al. Efficient multi-scale attention module with cross-spatial learning[J]. arXiv:2305.13563, 2023.
[29] BOLYA D, FOLEY S, HAYS J, et al. TIDE: a general toolbox for identifying object detection errors[C]//Proceedings of the European Conference on Computer Vision. Cham: Springer, 2020: 558-573.