Open-source data alternatives for regenerating urban road fine-grained features

Abstract

The extraction of urban road features provides indispensable support to numerous high-accurate applications such as autonomous driving and urban high-definition mapping. However, approaches mainly focus on road connectivity, while often overlooking finer details of urban road constituent structures. Data that captures road details, such as LiDAR, may not be always readily available. This article proposes an operational framework for mapping fine-grained urban road features by integrating open-source data (OSD). The geometric measurement method is successively presented using projective geometry and prior knowledge for urban road sections. And a feature generation strategy is introduced to achieve and express the fine-grained road features. Compared with the corresponding large-scale topographic map (LTM) and available optical remote sensing image (AORSI), the proposed method regenerated fine-grained features of urban roads with m-level. It provides a cost-effective innovative and alternative method to acquiring fine-grained road datasets in other data-scarce regions.

Keywords

• Urban road
• fine-grained
• open-source data
• street view image
• projective geometry