Pavement distress data has been widely collected through pavement survey to evaluate roadway condition. Despite rapid advancements in automated pavement distress data collection, the implementation of the widely used pavement condition index (PCI) still requires extensive manual labor either in tradition manual survey or image-based PCI inspections. This paper proposes a deep-learning (DL) based automatic pavement cracking detection algorithm and further evaluates a potential for automatic PCI survey with the obtained cracking results on asphalt pavement. The pavement two- and three-dimensional (2D and 3D) images are collected via the state-of-the-art PaveVision3D imaging system on two asphalt sites in Maryland. The automatic lane marking detection is performed on the 2D images to define the lane width for the following cracking detection. The DL based CrackNet automates the detection of the cracking information from the 3D images. Cracking information for alligator, longitudinal, and transverse cracking are summarized to automatically calculate the PCI numbers. The high correlations between the fully automated PCI results and the historical PCI results indicate the DL based automatic PCI calculation method produces reasonable PCI numbers with minimum human intervention.
Field Performance of Deep-Learning Based Fully Automated Cracking Analysis and Its Potential for PCI Surveys
International Airfield and Highway Pavements Conference 2019 ; 2019 ; Chicago, Illinois
2019-07-18
Conference paper
Electronic Resource
English
Precision Test of Cracking Surveys with the Automated Distress Analyzer
Online Contents | 2011
|Precision Test of Cracking Surveys with the Automated Distress Analyzer
British Library Online Contents | 2011
|Online Contents | 1993
|Elsevier | 1993
|British Library Conference Proceedings | 1993
|