Detection of Damaged Road Paints of Crosswalks by Focusing on Multi-layered Features

Abstract

In this study, we aim to develop a detection system of damaged crosswalks as a basic component of a digital map localization system. In rural areas, because the road paints of crosswalks are sometimes damaged, the features on which the existing methods focus for detecting them, such as rectangular shapes with side edges, are not clear. Thus, we focus on multi-layered faint features: existence of the white-band bottom shape, distribution of the white-band bottoms, and shape of the white-band candidates. Through an experiment on public roads, we confirm the practical performance of the proposed system.