|
|
基于改进ViT-YOLOv5的高速公路表面裂缝检测研究
|
Abstract:
道路裂缝会对行车安全构成威胁甚至引发交通事故，现有的检测方法在精度和准确率上仍然有改进的空间。因此提出了一种基于改进的ViT-YOLOv5 (Transformer-You Only Look Once Version 5)的高速公路表面裂缝检测研究。首先引入了双向特征金字塔网络(BiFPN)特征融合结构，并引入了自注意力机制和改进的Squeeze-and-Excitation (SE)注意力机制，动态地调整每个通道的特征权重，对输入特征图的不同通道给予不同的重视程度。其次借鉴了BiFPN的双向跨尺度连接和加权特征融合结构，改进了YOLOv5的头部网络，更有效地实现了多尺度特征融合。最后采用Transformer的核心思想，设计了Transformer层模块，引入了自注意力机制，增强了模型关注信息的不同方面的能力。实验结果表明，改进后的目标检测算法在高速公路表面裂缝检测中的平均精度mAP@0.5，与传统YOLOv5算法相比提升了28个百分点。改进后的算法能够快速、准确地提高高速公路表面裂缝检测精度。
Serious freeway pavement crack can pose a threat to driving safety and even cause traffic accidents. Existing detection methods still have room for improvement in precision and accuracy. Therefore a method on freeway surface crack detection based on the improved ViT-YOLOv5 (Transformer-You Only Look Once Version 5) is proposed. Firstly, the feature fusion structure of bidirectional feature pyramid network (BiFPN) feature is introduced, and the self-attention mechanism and the im-proved Squeeze-and-Excitation (SE) attention mechanism are introduced to dynamically adjust the feature weight of each channel and give different attention to different channels of the input feature map. Secondly, by drawing on the bidirectional cross-scale connection and weighted feature fusion structure of BiFPN, the head network of YOLOv5 is improved to realize multi-scale feature fusion more effectively. Finally, the Transformer layer module is designed, based on the core idea of Transformer, and the self-attention mechanism is introduced, which enhances the ability of the model to pay attention to different aspects of information. The results show that the average accu-racy of the improved object detection algorithm in freeway surface crack detection is mAP@0.5, which is 28 percentage more than the traditional YOLOv5 algorithm. The improved algorithm can quickly and accurately improve the accuracy of highway surface crack detection.
| [1] | 常州大学. 一种基于改进YOLO的道路裂缝检测方法及系统[P]. 中国, CN202210854022.9. 2022-10-18. |
| [2] | 王骜. 基于图像特征的路面破损检测方法及装备技术[D]: [硕士学位论文]. 武汉: 武汉工程大学, 2016. |
| [3] | 王勇. 基于车载360？环视影像和深度神经网络的路面坑塘与裂缝检测[J]. 电气自动化, 2020, 42(4): 99-102. |
| [4] | 张志华, 温亚楠, 慕号伟, 等. 结合双注意力机制的道路裂缝检测[J]. 中国图象图形学报, 2022, 27(7): 2240- 2250. |
| [5] | Lekshmipathy, J., Samuel, N.M. and Velayudhan, S. (2020) Vibration vs. Vision: Best Approach for Automated Pavement Distress Detection. International Journal of Pavement Research and Technology, 13, 402-410.
https://doi.org/10.1007/s42947-020-0302-y |
| [6] | Zhang, Y., Ma, Z., Song, X., Wu, J., Liu, S., Chen, X. and Guo, X. (2021) Road Surface Defects Detection Based on IMU Sensor. IEEE Sensors Journal, 22, 2711-2721. https://doi.org/10.1109/JSEN.2021.3135388 |
| [7] | Dorafshan, S., Thomas, R.J. and Maguire, M. (2018) Comparison of Deep Convolutional Neural Networks and Edge Detectors for Image-Based Crack Detection in Concrete. Construction and Building Materials, 186, 1031-1045.
https://doi.org/10.1016/j.conbuildmat.2018.08.011 |
| [8] | Peng, C., Yang, M.Q., Zheng, Q.H., Zhang, J., Wang, D.Q., Yan, R.Y., Wang, J.X. and Li, B.J. (2020) A Triple- Thresholds Pavement Crack Detection Method Leveraging Random Structured Forest. Construction and Building Materials, 263, Article ID: 120080. https://doi.org/10.1016/j.conbuildmat.2020.120080 |
| [9] | Amhaz, R., Chambon, S., Idier, J. and Baltazart, V. (2016) Au-tomatic Crack Detection on Two-Dimensional Pavement Images: An Algorithm Based on Minimal Path Selection. IEEE Trans-actions on Intelligent Transportation Systems, 17, 2718-2729. https://doi.org/10.1109/TITS.2015.2477675 |
| [10] | F. Road-ware. https://www.fugro.com/our-services/asset-integrity/roadware |
| [11] | Park, S.E., Eem, S.-H. and Jeon, H. (2020) Concrete Crack Detection and Quantification Using Deep Learning and Structured Light. Construction and Building Materials, 252, Article ID: 119096.
https://doi.org/10.1016/j.conbuildmat.2020.119096 |
| [12] | Tong, Z., Gao, J. and Zhang, H. (2018) Innovative Method for Recognizing Subgrade Defects Based on a Convolutional Neural Network. Construction and Building Materials, 169, 69-82.
https://doi.org/10.1016/j.conbuildmat.2018.02.081 |
| [13] | Mao, Z.J., Zhao, C.H., Zheng, Y.F., Mao, Y., Li, H., Hua, L.R. and Liu, Y. (2020) Research on Detection Method of Pavement Diseases Based on Unmanned Aerial Vehicle (UAV). Interna-tional Conference on Image, Video Processing and Artificial Intelligence, Volume 11584, 127-132. https://doi.org/10.1117/12.2580285 |
| [14] | Simonyan, K. and Zisserman, A. (2014) Very Deep Convolutional Networks for Large-Scale Image Recognition. 1-14. |
| [15] | Opara, J.N., Thein, A.B.B., Izumi, S., Yasuhara, H. and Chun, P.J. (2021) Defect Detection on Asphalt Pavement by Deep Learning. International Journal of GEOMATE, 21, 87-94. https://doi.org/10.21660/2021.83.6153 |
| [16] | Arya, D., Maeda, H., Ghosh, S.K., Toshniwal, D., Mraz, A., Kashiyama, T. and Sekimoto, Y. (2021) Deep Learning-Based Road Damage Detection and Classification for Multiple Countries. Automation in Construction, 132, Article ID: 103935 https://doi.org/10.1016/j.autcon.2021.103935 |
| [17] | Howard, A.G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M. and Adam, H. (2017) MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications. 1-9. |
| [18] | Hegde, V., Trivedi, D., Alfarrarjeh, A., Deepak, A., Kim, S.H. and Shahabi, C. (2020) Yet Another Deep Learning Approach for Road Damage Detection Using Ensemble Learning. 2020 IEEE International Conference on Big Data (Big Data), Atlanta, 10-13 December 2020, 5553-5558. https://doi.org/10.1109/BigData50022.2020.9377833 |
| [19] | Doshi, K. and Yilmaz, Y. (2020) Road Damage Detection Us-ing Deep Ensemble Learning. 2020 IEEE International Conference on Big Data (Big Data), Atlanta, 10-13 December 2020, 5540-5544.
https://doi.org/10.1109/BigData50022.2020.9377774 |
| [20] | Bochkovskiy, A., Wang, C.-Y. and Liao, H.-Y.M. (2020) YOLOv4: Optimal Speed and Accuracy of Object Detection. 1-17. |
| [21] | Pei, Z., Lin, R., Zhang, X., Shen, H., Tang, J. and Yang, Y. (2020) CFM: A Consistency Filtering Mechanism for Road Damage Detection. 2020 IEEE International Conference on Big Data (Big Data), Atlanta, 10-13 December 2020, 5584-5591. https://doi.org/10.1109/BigData50022.2020.9377911 |
| [22] | Cai, Z. and Vasconcelos, N. (2019) CascadeR-CNN: High Quality Object Detection and Instance Segmentation.
https://doi.org/10.1109/CVPR.2018.00644 |
