Hybrid Procedure for Automated Detection of Cracking with 3D Pavement Data

Pavement cracks are considered a major indicator of pavement performance. Because traditional manual crack surveys are dangerous, time consuming, and expensive, technologies have been developed to collect high-speed pavement images, and numerous algorithms have been proposed to detect cracks on pavement surface. The latest PaveVision3D Ultra system (3D Ultra) has been implemented to achieve 30-kHz three-dimensional (3D) scanning rate for 1-mm resolution pavement surface data at highway speed up to 100 km/h (60 mi/h). This paper presents the application of a hybrid procedure for automated crack detection on 3D pavement data collected using 3D Ultra. The procedure combines three different methods, namely, the matched filtering (MF) to highlight the cracks, the tensor voting to determine the main directions of the cracks, and the minimum spanning tree to identify the crack paths. The authors provide comparisons with cracking-detection results from traditional edge detectors and reference crack maps produced by a semiautomated software. The experimental results, through performance measurements and a statistical point of view, show that the proposed algorithm is able to detect pavement cracks with high precision. Finally, this paper also discusses preliminary considerations for exploiting the MF features to evaluate the orientation of the various crack segments.