基于改进YOLOv5的路面病害检测模型

doi:10.19678/j.issn.1000-3428.0064924

摘要/Abstract

摘要： 针对路面病害检测中由于病害形态多样、种类繁多以及背景灰度值相似造成噪声干扰导致识别与分类精度不高的问题，采用卷积神经网络YOLOv5为主干框架，提出一种基于改进YOLOv5的路面病害检测模型YOLOv5l-CBF。引入坐标注意力机制，调整网络的注意力权重使模型对病害纹理特征更加关注，并在主干网络的残差结构中引入Transformer构建BotNet网络结构，在减少参数量的同时提高对病害图像中全局依赖关系的捕捉能力。同时，在颈部网络中构建双向加权特征金字塔网络，学习每个特征层的重要性分布权重，并对提取到的病害特征进行双向交叉尺度连接和加权融合。在真实路面病害数据集上的实验结果表明：与YOLOv5l模型相比，YOLOv5l-CBF模型精度与召回率分别提升7.4和8.7个百分点，mAP达到90.8%，在对多种病害的检测与分类上具有显著的性能优势。

关键词: 目标检测, 路面病害, 注意力机制, 双向加权特征融合, 卷积神经网络

Abstract: To address the problem of poor recognition and classification accuracy in pavement disease detection due to the diverse morphologies, the types of diseases, and the interference of similar background gray scale values, a Convolutional Neural Network(CNN) YOLOv5 is used as the basic framework, and pavement disease detection model YOLOv5l-CBF based on improved YOLOv5 is proposed.The Coordinate Attention(CA) mechanism is introduced in the network, and the attention weight of the network is adjusted to focus on the texture features of the disease.A Transformer is introduced into the residual structure of the backbone network to construct the BotNet network structure, which can reduce the number of parameters and improve the ability to capture globally dependent in disease images.A Bi-directional weighted Feature Pyramid Network(BiFPN) is constructed at the Neck, such that the network detects the important distribution weights of each feature layer, and the extracted disease features will achieve effective bi-directional cross-scale connection and weighted fusion.Experimental results on an actual pavement disease dataset demonstrate that compared with YOLOv5l model, the YOLOv5l-CBF model improves the accuracy and recall of disease identification and classification by 7.4 and 8.7 percentage points, respectively, and the mean Average Precision(mAP) can reach 90.8%.It has significant advantages for detecting and classifying multiple disease types in highway environments.

Key words: object detection, pavement disease, attention mechanism, bidirectional weighted feature fusion, Convolutional Neural Network(CNN)

中图分类号:

TP391

王朕, 李豪, 严冬梅, 竺永荣. 基于改进YOLOv5的路面病害检测模型[J]. 计算机工程, 2023, 49(2): 15-23.

WANG Zhen, LI Hao, YAN Dongmei, ZHU Yongrong. Pavement Disease Detection Model Based on Improved YOLOv5[J]. Computer Engineering, 2023, 49(2): 15-23.

https://www.ecice06.com/CN/Y2023/V49/I2/15

图/表 13

20230216175703

20230216175707

20230216175710

20230216175714

20230216175812

20230216175816

20230216175819

20230216175823

20230216175826

20230216175829

20230216175832

20230216175836

20230216175857

参考文献

[1] DOSHI K, YILMAZ Y.Road damage detection using deep ensemble learning[C]//Proceedings of IEEE International Conference on Big Data.Washington D.C., USA:IEEE Press, 2020:5540-5544.
[2] 王程, 刘元盛, 刘圣杰.基于改进YOLOv4的小目标行人检测算法[J/OL].计算机工程:1-9[2022-05-01].DOI:10.19678/j.issn.1000-3428.0063623. WANG C, LIU Y S, LIU S J.Small target detection algorithm based on improved YOLOv4 for pedestrian[J/OL].Computer Engineering:1-9[2022-05-01].DOI:10.19678/j.issn.1000-3428.0063623.(in Chinese)
[3] TACK A, PREIM B, ZACHOW S.Fully automated assessment of knee alignment from full-leg x-rays employing a "YOLOv4 and Resnet landmark regression algorithm"(YARLA):data from the osteoarthritis initiative[J].Computer Methods and Programs in Biomedicine, 2021, 205:106080.
[4] GIRSHICK R, DONAHUE J, DARRELL T, et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2014:580-587.
[5] GIRSHICK R.Fast R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2015:1440-1448.
[6] REN S Q, HE K M, GIRSHICK R, et al.Faster R-CNN:towards real-time object detection with region proposal networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6):1137-1149.
[7] HE K M, GKIOXARI G, DOLLÁR P, et al.Mask R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2017:2980-2988.
[8] REDMON J, DIVVALA S, GIRSHICK R, et al.You only look once:unified, real-time object detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2015:779-788.
[9] LIU W, ANGUELOV D, ERHAN D, et al.SSD:single shot multi-box detector[C]//Proceedings of European Conference on Computer Vision.Berlin, Germany:Springer, 2015:21-37.
[10] GOU C, PENG B, LI T R, et al.Pavement crack detection based on the improved Faster-RCNN[C]//Proceedings of the 14th International Conference on Intelligent Systems and Knowledge Engineering.Washington D.C., USA:IEEE Press, 2019:962-967.
[11] 章世祥, 张汉成, 李西芝, 等.基于机器视觉的路面裂缝病害多目标识别研究[J].公路交通科技, 2021, 38(3):30-39. ZHANG S X, ZHANG H C, LI X Z, et al.Study on multi-objective identification of pavement cracks based on machine vision[J].Journal of Highway and Transportation Research and Development, 2021, 38(3):30-39.(in Chinese)
[12] HUANG R, PEDOEEM J, CHEN C X.YOLO-LITE:a real-time object detection algorithm optimized for non-GPU computers[C]//Proceedings of IEEE International Conference on Big Data.Washington D.C., USA:IEEE Press, 2018:2503-2510.
[13] NIE M X, WANG C.Pavement crack detection based on yolo v3[C]//Proceedings of the 2nd International Conference on Safety Produce Informatization.Washington D.C., USA:IEEE Press, 2019:327-330.
[14] FARAMARZI M.Road damage detection and classification using deep neural networks(YOLOv4) with smartphone images[EB/OL].[2022-05-01].https://ssrn.com/abstract=3627382.
[15] YAN K, ZHANG Z H.Automated asphalt highway pavement crack detection based on deformable single shot multi-box detector under a complex environment[J].IEEE Access, 2021, 9:150925-150938.
[16] SONG K, ZHOU X Q, YU H, et al.Towards better word alignment in Transformer[J].ACM Transactions on Audio, Speech, and Language Processing, 2020, 28:1801-1812.
[17] DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al.An image is worth 16×16 words:Transformers for image recognition at scale[EB/OL].[2022-05-01].https://arxiv.org/abs/2010.11929.
[18] LIU Z, LIN Y T, CAO Y, et al.Swin Transformer:hierarchical vision Transformer using shifted windows[C]//Proceedings of IEEE/CVF International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2021:9992-10002.
[19] HOU Q B, ZHOU D Q, FENG J S.Coordinate attention for efficient mobile network design[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2021:13708-13717.
[20] TAN M X, PANG R M, LE Q V.EfficientDet:scalable and efficient object detection[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2020:10778-10787.
[21] MAHENDRAKAR T, WHITE R T, WILDE M, et al.Real-time satellite component recognition with YOLOv5[C]//Proceedings of Small Satellite Conference.Logan, USA:[s.n.], 2021:1-12.
[22] LONG X, DENG K, WANG G, et al.PP-YOLO:an effective and efficient implementation of object detector[EB/OL].[2022-05-01].https://arxiv.org/abs/2007.12099.
[23] CHENG S, WANG L, DU A.Asymmetric coordinate attention spectral-spatial feature fusion network for hyperspectral image classification[J].Scientific Reports, 2021, 11(1):1-17.
[24] ZHU L L, GENG X, LI Z, et al.Improving YOLOv5 with attention mechanism for detecting boulders from planetary images[J].Remote Sensing, 2021, 13(18):3776.
[25] LIN T Y, DOLLÁR P, GIRSHICK R, et al.Feature pyramid networks for object detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2017:936-944.
[26] 交通运输部公路科学研究院.公路技术状况评定标准:JTG 5210-2018[S].北京:人民交通出版社, 2018. Institute of Highway Science, Ministry of Transport.Highway technical condition evaluation standard:JTG 5210-2018[S].Beijing:People's Communications Press, 2018.(in Chinese).

选择文件类型/文献管理软件名称

选择包含的内容