• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学 ›› 2022, Vol. 44 ›› Issue (04): 674-685.

• 图形与图像 • 上一篇    下一篇

基于可见光视觉图像的路面裂缝识别深度学习方法述评

卢凯良   

  1. (江苏自动化研究所,江苏 连云港 222061)
  • 收稿日期:2020-12-28 修回日期:2021-03-22 接受日期:2022-04-25 出版日期:2022-04-25 发布日期:2022-04-20

Advances in deep learning methods for pavement crack detection and identification with visual images

LU Kai-liang   

  1. (Jiangsu Automation Research Institute,Lianyungang 222061,China)

  • Received:2020-12-28 Revised:2021-03-22 Accepted:2022-04-25 Online:2022-04-25 Published:2022-04-20

摘要: 基于可见光视觉图像的表面裂缝识别为非接触式,不受被测对象材质限制,可在线自动检测,具有速度快、成本低和精度高等优势。首先较为全面地搜集了典型的路面裂缝公开数据集,整理归纳了样本特征及其随机可变影响因素,并比较了传统手工设计特征工程、机器学习和深度学习3种主要裂缝识别方法的优缺点。然后,从网络架构、性能和效果方面着重评述了自搭架构、迁移学习和编码-解码器等易于训练和部署的深度学习算法新进展,通过算法优化和算力提升可显著提高识别的效果和性能,测试结果表明能够在低算力平台上实现裂缝补丁级快速检测和像素级实时检测。

关键词: 路面裂缝检测, 视觉图像识别, 深度学习

Abstract: Surface crack identification with visual images is a kind of non-contact detection solution, which is not limited by the material of the tested object and is easy to achieve online automation. Therefore, it has the advantages of fast speed, low cost and high precision. Firstly, the public data sets of typical pavement crack are comprehensively collected, and the characteristics of sample images and the random variable factors are summarized. Subsequently, the advantages and disadvantages of hand-crafted feature engineering, machine learning, and deep learning crack identification methods are compared. Finally, from the aspects of network architecture, testing performance and predicting effectiveness, this paper reviews the development and progress of typical deep learning algorithms such as self-built CNN, transfer learning (TL) and encoder-decoder (ED) that can be easily trained and deployed. We can see the obvious improvement of performance and effect because of algorithm optimization and computing power enhancement. The results show that real-time patch-level and fast pixel-level crack detection can be realized on low computing power GPU platform.

Key words: pavement crack detection, visual image identification, deep learning