Research article

Embedded system for road damage detection by deep convolutional neural network

  • Received: 02 June 2019 Accepted: 15 August 2019 Published: 03 September 2019
  • Road pavement could be damaged due to various reasons, causing damages such as cracks and pits. These damages cause potential dangers in traffic safety. It is necessary for road maintenance departments to find damages in time before maintenance. At present, maintenance departments of some high-level roads are equipped with specialized detection vehicles such as laser scanning vehicles to detect road damages. These kinds of devices can get good detection performance, but the economic cost is very high. In this paper, we use a road damage image dataset to train an object detection model based on deep convolutional neural network and deploy it on a low-cost embedded platform to form an embedded system. The system uses a common camera mounted on windshield of a common vehicle as sensor to detect road damages. The embedded system consumes about 352 ms to process one frame of image and can achieve a recall rate of about 76% which is higher than some previous related works. The recall rate of this scheme using common cameras is less than that of high-level specialized detectors, but the economic cost is much lower than them. After subsequent development, the road maintenance department with limited funds can consider about schemes like this.

    Citation: Siyu Chen, Yin Zhang, Yuhang Zhang, Jiajia Yu, Yanxiang Zhu. Embedded system for road damage detection by deep convolutional neural network[J]. Mathematical Biosciences and Engineering, 2019, 16(6): 7982-7994. doi: 10.3934/mbe.2019402

    Related Papers:

  • Road pavement could be damaged due to various reasons, causing damages such as cracks and pits. These damages cause potential dangers in traffic safety. It is necessary for road maintenance departments to find damages in time before maintenance. At present, maintenance departments of some high-level roads are equipped with specialized detection vehicles such as laser scanning vehicles to detect road damages. These kinds of devices can get good detection performance, but the economic cost is very high. In this paper, we use a road damage image dataset to train an object detection model based on deep convolutional neural network and deploy it on a low-cost embedded platform to form an embedded system. The system uses a common camera mounted on windshield of a common vehicle as sensor to detect road damages. The embedded system consumes about 352 ms to process one frame of image and can achieve a recall rate of about 76% which is higher than some previous related works. The recall rate of this scheme using common cameras is less than that of high-level specialized detectors, but the economic cost is much lower than them. After subsequent development, the road maintenance department with limited funds can consider about schemes like this.


    加载中


    [1] E. Chuo, Research history and prospect of domestic pavement automatic detection system, China High-Tech. Enterp., 1 (2011), 1-3.
    [2] G. E. Hinton, S. Osindero and Y. W. Teh, A fast learning alg Comput., 18 (2006), 1527-1554. rithm for deep belief nets, Neural
    [3] L. Zhang, F. Yang, Y. D. Zhang, et al., Road crack detection using deep convolutional neural network, 2016 IEEE Int. Conf. Image Process. (ICIP), 3708-3712.
    [4] H. Maeda, Y. Sekimoto, T. Seto, et al., Road damage detection using deep neural networks with images captured through a smartphone, Comput. Aided Civ. Infrastruct. Eng., 33 (2018), 1127-1141. doi: 10.1111/mice.12387
    [5] W. Liu, D. Anguelov, D. Erhan, et al., SSD:Single shot multiBox detector, 2016 European Conf. Comput. Vision (ECCV), 21-37.
    [6] A. G. Howard, M. Zhu, B. Chen, et al., MobileNets:Efficient convolutional neural networks for mobile vision applications, preprint, arXiv:1704.04861.
    [7] A. He, Advantages of JG-1 laser 3D pavement inspection system, China Highw., 16 (2005), 94-95.
    [8] R. Girshick, J. Donahue, T. Darrell, et al., Rich feature hierarchies for accurate object detection and semantic segmentation, 2014 IEEE Conf. Comput. Vision Pattern 587. Recogn. (CVPR), 580-587
    [9] S. Ren, K. He, R. Girshick, et al., Faster R-CNN:Towards real-time object detection with region Proposal networks, Adv. Neural Inf. Process. Syst., (2015), 91-99.
    [10] J. Redmon, S. Divvala, R. Girshick, et al., You only look once:Unified, Real-Time object detection, 2016 IEEE Conf. Comput. Vision Pattern Recogn. (CVPR), 779-788.
    [11] R. Joseph and A. Farhadi, YOLO v3:An incremental improvement, preprint, arXiv:1804.02767.
    [12] S. Karen and A. Zisserman, Very recognition, preprint, arXiv:1409.1556. deep convolutional networks for Large-Scale image
    [13] Z. Zhou, Model evaluating and selecting, in Machine Learn ng (eds. Zhihua Zhou), Tsinghua University Press, (2016), 23-52.
    [14] T. Lin, M. Maire, S. Belongie, et al., Microsoft COCO:Common objects in context, 2014 European Conf. Comput. Vision (ECCV), 740-755.
  • Reader Comments
  • © 2019 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(2044) PDF downloads(624) Cited by(2)

Article outline

Figures and Tables

Figures(10)  /  Tables(2)

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog