Mapping in the Future: Advancing HD Maps Creation with Semi-Automated Feature Extraction

Abstract. The production of high-definition maps (HD Maps) is a multi-stage, resource-intensive process that demands substantial investments in specialized equipment, skilled labor, and time. This study introduces a semi-automated mapping tool aimed at addressing these challenges through the integration of point cloud data, trajectory information, and image-based AI algorithms. One of the key innovations of this tool is a user-friendly graphical user interface (GUI), which enhances usability by facilitating data import, preprocessing customization, and feature visualization. The tool focuses on extracting essential road features such as lane lines, stop lines, directional arrows, and traffic signals, outputting data in various formats including LAS, PCD, and SHP. Performance evaluations were conducted in both controlled and real-world environments. In the Taiwan CARLab, the tool demonstrated high accuracy under diverse traffic scenarios. Testing on Taiwan's National Highway No. 1 further confirmed the tool’s robustness in handling real-world conditions, achieving up to a 50–70% reduction in processing time compared to manual digitization. These findings highlight the tool's potential to significantly reduce production costs while maintaining accuracy, thereby facilitating wider adoption of HD Maps in autonomous driving applications.