Vision-based multi-perspective road pre-damage detection

Abstract

Road condition evaluation is essential for selecting maintenance strategies to prevent road failures. However, current road evaluation methods are subjective, manually performed and time-consuming, while road maintenance is mostly reactive with associated risks, such as possible network disruptions. Presented herein is a vision-based methodology for road inspection and evaluation. For this purpose, we couple digital image correlation (DIC) with perspective warping to measure material displacement or strain that can be an early warning of pavement failure. Our results show that perspective warped images were able to be successfully used in DIC analysis and obtain strain values within 19% of the images taken at a direct angle to the sample. Perspective warping enables us to bypass the limitation of fixed DIC applications and allows for the collection of images from mobile cameras such as everyday cars in motion. This provides new opportunities for the effective prediction of road damage formation and consequently predictive maintenance.