In response to the issues of low efficiency and high cost in traditional manual methods for road surface crack detection, an improved YOLOv5s (you only look once version 5 small) algorithm was proposed. Based on this improvement, a road surface crack object recognition model was established using YOLOv5s. First, based on the Res2Net (a new multi-scale backbone architecture) network, an improved multi-scale Res2-C3 (a new multi-scale backbone architecture of C3) module was suggested to enhance feature extraction performance. Second, the feature fusion network and backbone of YOLOv5 were merged with the GAM (global attention mechanism) attention mechanism, reducing information dispersion and enhancing the interaction of global dimensions features. We incorporated dynamic snake convolution into the feature fusion network section to enhance the model's ability to handle irregular shapes and deformation problems. Experimental results showed that the final revision of the model dramatically increased both the detection speed and the accuracy of road surface identification. The mean average precision (mAP) reached 93.9%, with an average precision improvement of 12.6% compared to the YOLOv5s model. The frames per second (FPS) value was 49.97. The difficulties of low accuracy and slow speed in road surface fracture identification were effectively addressed by the modified model, demonstrating that the enhanced model achieved relatively high accuracy while maintaining inference speed.
Citation: Jiaming Ding, Peigang Jiao, Kangning Li, Weibo Du. Road surface crack detection based on improved YOLOv5s[J]. Mathematical Biosciences and Engineering, 2024, 21(3): 4269-4285. doi: 10.3934/mbe.2024188
[1] | Nick Cercone . What's the Big Deal About Big Data?. Big Data and Information Analytics, 2016, 1(1): 31-79. doi: 10.3934/bdia.2016.1.31 |
[2] | Ali Asgary, Jianhong Wu . ADERSIM-IBM partnership in big data. Big Data and Information Analytics, 2016, 1(4): 277-278. doi: 10.3934/bdia.2016010 |
[3] | Yaguang Huangfu, Guanqing Liang, Jiannong Cao . MatrixMap: Programming abstraction and implementation of matrix computation for big data analytics. Big Data and Information Analytics, 2016, 1(4): 349-376. doi: 10.3934/bdia.2016015 |
[4] | Pankaj Sharma, David Baglee, Jaime Campos, Erkki Jantunen . Big data collection and analysis for manufacturing organisations. Big Data and Information Analytics, 2017, 2(2): 127-139. doi: 10.3934/bdia.2017002 |
[5] | Enrico Capobianco . Born to be Big: data, graphs, and their entangled complexity. Big Data and Information Analytics, 2016, 1(2): 163-169. doi: 10.3934/bdia.2016002 |
[6] | John A. Doucette, Robin Cohen . A testbed to enable comparisons between competing approaches for computational social choice. Big Data and Information Analytics, 2016, 1(4): 309-340. doi: 10.3934/bdia.2016013 |
[7] |
Hamzeh Khazaei, Marios Fokaefs, Saeed Zareian, Nasim Beigi-Mohammadi, Brian Ramprasad, Mark Shtern, Purwa Gaikwad, Marin Litoiu .
How do I choose the right NoSQL solution? A comprehensive theoretical and experimental survey . Big Data and Information Analytics, 2016, 1(2): 185-216.
doi: 10.3934/bdia.2016004
|
[8] | Richard Boire . UNDERSTANDING AI IN A WORLD OF BIG DATA. Big Data and Information Analytics, 2018, 3(1): 22-42. doi: 10.3934/bdia.2018001 |
[9] | M Supriya, AJ Deepa . Machine learning approach on healthcare big data: a review. Big Data and Information Analytics, 2020, 5(1): 58-75. doi: 10.3934/bdia.2020005 |
[10] | Weidong Bao, Wenhua Xiao, Haoran Ji, Chao Chen, Xiaomin Zhu, Jianhong Wu . Towards big data processing in clouds: An online cost-minimization approach. Big Data and Information Analytics, 2016, 1(1): 15-29. doi: 10.3934/bdia.2016.1.15 |
In response to the issues of low efficiency and high cost in traditional manual methods for road surface crack detection, an improved YOLOv5s (you only look once version 5 small) algorithm was proposed. Based on this improvement, a road surface crack object recognition model was established using YOLOv5s. First, based on the Res2Net (a new multi-scale backbone architecture) network, an improved multi-scale Res2-C3 (a new multi-scale backbone architecture of C3) module was suggested to enhance feature extraction performance. Second, the feature fusion network and backbone of YOLOv5 were merged with the GAM (global attention mechanism) attention mechanism, reducing information dispersion and enhancing the interaction of global dimensions features. We incorporated dynamic snake convolution into the feature fusion network section to enhance the model's ability to handle irregular shapes and deformation problems. Experimental results showed that the final revision of the model dramatically increased both the detection speed and the accuracy of road surface identification. The mean average precision (mAP) reached 93.9%, with an average precision improvement of 12.6% compared to the YOLOv5s model. The frames per second (FPS) value was 49.97. The difficulties of low accuracy and slow speed in road surface fracture identification were effectively addressed by the modified model, demonstrating that the enhanced model achieved relatively high accuracy while maintaining inference speed.
Researchers in the computational intelligence society have been consistently achieving progress in making machines more intelligent from various aspects, including representations, learning models, and optimization methods. The development of these techniques provides useful tools for big data and information analytics. This special issue aims at presenting recent advancements of combining computational intelligence methods with big data. We accepted 7 papers after a strict review process. Each paper was reviewed by at least two reviewers. We hope the accepted papers to this special issue will provide a useful reference for researchers who are interested in computational intelligence and big data, and inspire more possibilities of novel methods and applications.
The accepted papers can be roughly divided into three categories, according to the aspects they involve.
On the aspect of representation, the article "Multiple-instance learning for text categorization based on semantic representation" by Zhang et al. employs the multi-instance representation for text data. A text document is usually represented as a single feature vector, which could be insufficient to expose its rich content for learning. This paper, based on the popular word2vec technique, represents a document by multiple instances. In such a way, the semantic meanings of a document can be well exposed, and the experiments show improved performance over single instance representation.
Another article "A comparative study of robustness measures for cancer signaling networks" by Zhou et al. studies the cancer signaling data represented as a network. The information exchange pathways in the cancer signaling network are essential to the cure of cancer, thus it is meaningful to find a sensitive measure of the network that is highly correlated with patient survivability. This work investigates the robustness of 14 typical cancer signaling networks. Experiments find out that the natural connectivity is a promising measurement, which could be expected to help cancer treatments.
On the aspect of learning models, the extreme learning machine is a recently emerged simple neural network model with randomly determined connection weights. The article "Two-hidden-layer extreme learning machine based wrist vein recognition system" by Yue et al. employs such neural network with two hidden layers to achieve a good performance in the wrist vein recognition task with a satisfactory training time.
Incremental ability of learning models are often appealing. The article "Selective further learning of hybrid ensemble for class imbalanced Increment learning" by Lin and Tang addresses the class imbalance issue which naturally arises in incremental learning, and proposes an ensemble-based method Selective Further Learning, where different component learners handle different issues of the learning. Experiments show that the proposed method outperforms some recent state-of-the-art approaches.
On the aspect of optimization methods, the article "A clustering based mate selection for evolutionary optimization" by Zhang et al. introduces the mate selection mechanism into evolutionary algorithms. Helped by the clustering, the mate of an individual is restricted in the same cluster. With this new mechanism, the evolutionary algorithm optimizes a set of benchmark functions better.
Optimization is also related with representation. In the article "A moving block sequence-based evolutionary algorithm for resource investment project scheduling problems" by Yuan et al. proposes the moving block sequence representation for the resource investment project scheduling problem. The new representation can guarantee some good properties of the solved solution, and consequently the proposed approach shows superior performance on 450 benchmark instances.
Better optimization can lead to better learning. In the article "An evolutionary multiobjective method for low-rank and sparse matrix decomposition" by Wu et al, a multiobjective evolutionary approach is employed to solve the low-rank matrix decomposition problem. The multiobjective approach can well trade-off between low-rank and sparse objectives, leading to satisfied results on nature image analysis.
We thank all the authors for their contributions to this special issue, and the reviewers for their careful and insightful reviews. We also thank Prof. Jianhong Wu and Prof. Zongben Xu, the Editor-in-Chiefs of the Big Data and Information Analytics journal, and Prof. Zhi-Hua Zhou from the Editorial Board of the journal for the full support of this special issue, and the Aimsciences staff for managing this special issue.
[1] |
C..P. Meng, J. P. Li, J. Guo, C. L. Li, Analysis of common problems in ecological impact investigation of highway environmental protection acceptance, Res. Cons. Envir. Prot., 4 (2023), 121–124. https://doi.org/10.16317/j.cnki.12-1377/x.2023.04.015 doi: 10.16317/j.cnki.12-1377/x.2023.04.015
![]() |
[2] | W. Zhou, Y. He, J. Li, Dangerous behavior detection in gas stations based on deep learning, in 2023 IEEE 6th International Conference on Electronic Information and Communication Technology, (2023), 935–939. https://doi.org/10.1109/ICEICT57916.2023.10245093 |
[3] |
N. Sholevar, A. Golroo, S. R. Esfahani, Machine learning techniques for pavement condition evaluation, Autom. Constr., 136 (2022), 104190. https://doi.org/10.1016/j.autcon.2022.104190 doi: 10.1016/j.autcon.2022.104190
![]() |
[4] | H. Bello-Salau, A. M. Aibinu, E. N. Onwuka, J. J. Dukiya, A. J. Onumanyi, Image processing techniques for automated road defect detection: A survey, in International Conference on Electronics, (2014), 1–4. https://doi.org/10.1109/ICECCO.2014.6997556 |
[5] | S. Chatterjee, P. Saeedfar, S. Tofangchi, L. M. Kolbe, Intelligent road maintenance: a machine learning approach for surface defect detection, in European Conference on Information Systems, 2018. |
[6] | J. Redmon, S. Divvala, R. Girshick, A. Farhadi, You only look once: Unified, real-time object detection, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, (2016), 779–788. |
[7] |
S. Park, S. Bang, H. Kim, H. Kim, Patch-based crackdetection in black box images using convolutional neural net-works, J. Comput. Civil Eng., 33 (2019). https://doi.org/10.1061/(ASCE)CP.1943-5487.0000831 doi: 10.1061/(ASCE)CP.1943-5487.0000831
![]() |
[8] |
X. B. Su, Research on pavement crack detection based on improved YOLOv4, Henan Sci. Technol., 41 (2022), 62–67. https://doi.org/10.19968/j.cnki.hnkj.1003-5168.2022.18.012 doi: 10.19968/j.cnki.hnkj.1003-5168.2022.18.012
![]() |
[9] |
M. M. Wang, Q. D. Huang, S. N. Liu, Pavement damage detection based on improved YOLOv5s, J. Lasers, 44 (2023), 66–71. https://doi.org/10.14016/j.cnki.jgzz.2023.05.066 doi: 10.14016/j.cnki.jgzz.2023.05.066
![]() |
[10] | J. Terven, D. Cordova-Esparza, A comprehensive review of YOLO: From YOLOv1 to YOLOv8 and beyond, preprint, arXiv: 2304.00501. |
[11] |
J. Lu, M. Zhu, X. Ma, K. Wu, Steel strip surface defect detection Metho based on improved YOLOv5s, Biomimetics, 9 (2024), 28. https://doi.org/10.3390/biomimetics9010028 doi: 10.3390/biomimetics9010028
![]() |
[12] | Y. Zhou, W. Zhu, Y. He, Y. Li, Yolov8-based spatial target part recognition, in 2023 IEEE 3rd International Conference on Information Technology, Big Data and Artificial Intelligence, (2023), 1684–1687. https://doi.org/10.1109/ICIBA56860.2023.10165260 |
[13] |
H. Liu, F. Sun, J. Gu, L. Deng, Sf-yolov5: A lightweight small object detection algorithm based on improved feature fusion mode, Sensors, 22 (2022), 5817. https://doi.org/10.3390/s22155817 doi: 10.3390/s22155817
![]() |
[14] |
J. Zhou, Z. Xi, S. Wang, B. Yang, Y. Zhang, Y. Zhang, A real spatial–temporal attention denoising network for nugget quality detection in resistance spot weld, J. Intell. Manuf., 2023 (2023), 1–22. https://doi.org/10.1007/s10845-023-02160-x doi: 10.1007/s10845-023-02160-x
![]() |
[15] | C. Y. Wang, H. Y. M. Liao, Y. H. Wu, P. Y. Chen, J. W. Hsieh, IH Yeh CSPNet: A new backbone that can enhance learning capability of CNN, in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops, (2020), 390–391. |
[16] | K. Simonyan, A. Zisserman, Very deep convolutional networks for large-scale image recognition, preprint, arXiv: 1409.1556. |
[17] | C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, et al., Going deeper with convolutions, in Proceedings of the IEEE conference on computer vision and pattern recognition, (2015), 1–9. |
[18] |
X. Jiang, H. Hu, Y. Qin, Y. Hu, R. Ding, A real-time rural domestic garbage detection algorithm with an improved YOLOv5s network model, Sci. Rep., 12 (2022), 16802. https://doi.org/10.1038/s41598-022-20983-1 doi: 10.1038/s41598-022-20983-1
![]() |
[19] |
S. H. Gao, M. M. Cheng, K. Zhao, X. Y. Zhang, M. H. Yang, P. Torr, Res2Net: A New Multi-Scale Backbone Architecture, IEEE Trans. Pattern Anal. Mach. Intell., 2 (2021), 43. https://doi.org/10.1109/TPAMI.2019.2938758 doi: 10.1109/TPAMI.2019.2938758
![]() |
[20] |
U. Batool, M. I. Shapiai, S. A. Mostafa, M. Z. Ibrahim, An attention-augmented convolutional neural network with focal loss for mixed-type wafer defect classification, IEEE Access, 11 (2023), 108891–108905. https://doi.org/10.1109/ACCESS.2023.3321025 doi: 10.1109/ACCESS.2023.3321025
![]() |
[21] | Z. Yu, H. Huang, W. Chen, Y. Su, Y. Liu, X. Wang, Yolo-facev2: A scale and occlusion aware face detector, preprint, arXiv: 2208.02019. |
[22] |
J. Cai, J. Hu, 3D RANs: 3D residual attention networks for action recognition, Vis. Comput., 36 (2020), 1261–1270. https://doi.org/10.1007/s00371-019-01733-3 doi: 10.1007/s00371-019-01733-3
![]() |
[23] | J. Hu, L. Shen, G. Sun, Squeeze-and-Excitation networks, in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, (2018), 7132–7141. https://doi.org/10.1109/CVPR.2018.00745 |
[24] | S. Woo, J. Park, J. Y. Lee, I. S. Kweon, Cbam: Convolutional block attention module, in Proceedings of the European conference on computer vision (ECCV), (2018), 3–19. |
[25] | Y. Liu, Z. Shao, N. Hoffmann, Global attention mechanism: Retain information to enhance channel-spatial interactions, preprint, arXiv: 2112.05561. |
[26] | Z. Guo, Y. Li, Y. Tian, H. Liu, S. Yuan, C. Hou, Global attention-based approach for substation devices classification and localization, in 2023 IEEE/IAS Industrial and Commercial Power System Asia, (2023), 990–995. https://doi.org/10.1109/ICPSAsia58343.2023.10294513 |
[27] | Y. Qi, Y. He, X. Qi, Y. Zhang, G. Yang, Dynamic snake convolution based on topological geometric constraints for tubular structure segmentation, in Proceedings of the IEEE/CVF International Conference on Computer Vision, (2023), 6070–6079. https://doi.org/10.1109/ICCV51070.2023.00558 |
[28] | J. Dai, H. Qi, Y. Xiong, Y. Li, G. Zhang, H. Hu, et al., Deformable convolutional networks, in Proceedings of the IEEE international conference on computer vision, (2017), 764–773. https://doi.org/10.1109/ICCV.2017.89 |
[29] |
Z. Wu, R. Xue, H. Li, Real-time video fire detection via modified YOLOv5 network model, Fire Technol., 58 (2022), 2377–2403. https://doi.org/10.1007/s10694-022-01260-z doi: 10.1007/s10694-022-01260-z
![]() |
[30] |
Y. Wang, G. Fu, A novel object recognition algorithm based on improved YOLOv5 model for patient care robots, Int. J. Hum. Robot., 19 (2022). https://doi.org/10.1142/S0219843622500104 doi: 10.1142/S0219843622500104
![]() |
[31] |
L. Shi, S. Zhao, W. Niu, A welding defect detection method based on multiscale feature enhancement and aggregation, Nond. Testing and Eval., (2023), 1–20. https://doi.org/10.1080/10589759.2023.2253494 doi: 10.1080/10589759.2023.2253494
![]() |
[32] | E. R. Daniel, Wildfire smoke detection with computer vision, preprint, arXiv: 2301.05070. |