A two-step sequential automated crack detection and severity classification process for asphalt pavements

ABSTRACT

Crack detection, identification and classification are essential steps in pavement management system. It helps the agency in determining the appropriate rehabilitation technique to be done on the pavement. Typically, crack detection is done by manually checking images captured from road survey however it is uneconomical and time-consuming. In this study, a two-step sequential automated process detecting cracks and classifying severity of asphalt pavements is proposed. Mask RCNN was used in detecting and identifying linear and fatigue cracks (with severity) while image processing was used in determining the severity of linear crack. The Mask RCNN model used 21,150 images for training, composed of longitudinal, transverse and fatigue cracks of different severity levels. In addition, 5,657 and 5,756 images were utilised for validation and testing, respectively. The results showed that the Mask RCNN model can be used in detecting and identifying pavement cracks with an average accuracy of 92.10% while image processing was found to have a correlation coefficient of 0.85 when compared with field measured crack width and an accuracy of 87.5% in classifying linear cracks severity. The proposed method was found to be a promising approach in detecting cracks and classifying the corresponding severity of asphalt pavements.

KEYWORDS:

• Pavement distress
• crack detection
• Mask R-CNN
• deep learning
• image processing

Acknowledgements

This study has been done thanks to the partial support of the Korea Agency for Infrastructure Technology Advancement (KAIA) grant funded by the Ministry of Land, Infrastructure and Transport (Grant 20POQW-B152342-02) and Sejong University.

Disclosure statement

No potential conflict of interest was reported by the authors.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

This work was supported by the Korea Agency for Infrastructure Technology Advancement (KAIA) grant funded by the Ministry of Land, Infrastructure and Transport (Grant 20POQW-B152342-02) and Sejong University.