📝 SummarXiv

Abstract

Existing lane-level simulation road network generation is labor-intensive, resource-demanding, and costly due to the need for large-scale data collection and manual post-editing. To overcome these limitations, we propose automatically generating high-precision simulated road networks in traffic scenario, an efficient and fully automated solution. Initially, real-world road street view data is collected through open-source street view map platforms, and a large-scale street view lane line dataset is constructed to provide a robust foundation for subsequent analysis. Next, an end-to-end lane line detection approach based on deep learning is designed, where a neural network model is trained to accurately detect the number and spatial distribution of lane lines in street view images, enabling automated extraction of lane information. Subsequently, by integrating coordinate transformation and map matching algorithms, the extracted lane information from street views is fused with the foundational road topology obtained from open-source map service platforms, resulting in the generation of a high-precision lane-level simulation road network. This method significantly reduces the costs associated with data collection and manual editing while enhancing the efficiency and accuracy of simulation road network generation. It provides reliable data support for urban traffic simulation, autonomous driving navigation, and the development of intelligent transportation systems, offering a novel technical pathway for the automated modeling of large-scale urban road networks.