A road surface reconstruction dataset for autonomous driving

This work presents a road surface reconstruction dataset named RSRD, which to the best of our knowledge, is the first large-scale and real-world dataset special for road surface reconstruction. Real-vehicle experiments are performed with our hardware platform. High-resolution road surface stereo images, point clouds and motion information are acquired. This dataset can be utilized for computer vision and point cloud processing applications aiming at reconstructing road surface for saft and comfortable intelligent driving. Our dataset represents a pioneering contribution toward promoting autonomous driving by road surface reconstruction. It may contribute to both research and applications in terms of (i) developing universal 3D vision methods like monocular depth estimation, stereo matching, and multi-view stereo; (ii) exploring point cloud processing and motion estimation algorithms for robots and vehicles; (iii) estimating road unevenness and friction from reconstructed road profile and texture thus benefiting vehicle safety and comfort control systems; (iv) road crack monitoring for pavement maintenance. More details about the dataset are described in our webpage: https://thu-rsxd.com/rsrd.