Digitalising Road Maintenance: A Novel Approach for the Preparation and Integration of 2D and 3D Road Data Through Image Shadow Removal and Point Cloud Densification

Abstract

The digitalisation of road maintenance aims to combine digital technologies and harness data to revolutionise the current system to improve the safety, costs and sustainability of road networks. But still, current road data suffers from various issues -- road surface images are covered in shadows and 3D point cloud data is sparse. Inspection of road defects becomes problematic in such conditions. This paper automates 2D and 3D road data preparation, which will play a critical role in these digital advancements. We propose a two-fold solution: a physics-based shadow removal algorithm targeting shadows cast by the surveying vehicle and a Camera-LiDAR fusion technique for increasing point cloud density, both focusing on improving defects' visibility. Our solution shows significant improvements in automating data preparation. Shadow removal clears vehicle shadows with minimal artefacts remaining, revealing crack details and substantially enhancing the performance of patch detection with over four times the precision compared to raw images. The densification of the point cloud shows a marked increase in road surface detail, making defects visible.