Automatic Detection and Assessment of Pavement Marking Defects with Street View Imagery at the City Scale

Abstract

Pavement markings could wear out before their expected service life expires, causing traffic safety hazards. However, assessing pavement-marking conditions at the city scale was a great challenge in previous studies. In this article, we advance the method of detecting and evaluating pavement-marking defects at the city scale with Baidu Street View (BSV) images, using a case study in Nanjing. Specifically, we employ inverse perspective mapping (IPM) and a deep learning-based approach to pavement-marking extraction to make efficient use of street-view imageries. In addition, we propose an evaluation system to assess three types of pavement-marking defects, with quantitative and qualitative results provided for each image. Factors causing pavement-marking defects are discussed by mapping the spatial distribution of pavement-marking defects at the city scale. Our proposed methods are conducive to pavement-marking repair operations. Beyond this, this article can contribute to smart urbanism development by creating a new road maintenance solution and ensuring the large-scale realization of intelligent decision-making in urban infrastructure management.

Keywords

• pavement-marking defects
• street-view images
• urban management
• smart cities