Research Objective No. 3 – Automatic Detection and Vectorization of Transport Infrastructure Elements from Raster Orthophotos

This research objective focuses on the development and testing of a tool for the automatic detection of selected transport infrastructure elements, including the determination of their spatial and elevation coordinates. The main outcome will be a testing interface for an algorithm designed to automate the processing of heterogeneous datasets used in creating digital environment models, achieving significant time and cost savings while improving quality and accuracy compared to manual processing. The resulting models will be applicable, for example, in traffic simulators and other applications using virtual terrain models. The approach is based on high-resolution raster orthophotos acquired through aerial photogrammetry and ground-based 3D scanning, with additional testing of alternative imaging technologies and low-quality input data to assess robustness and associated risks. Methodologically, the system detects specific pixel color patterns representing selected road infrastructure elements, clusters and identifies them within the image, and converts raster data into vector format. The innovative aspect lies in performing automated vectorization only on selected image regions containing key objects rather than processing the entire raster image. The algorithm will be calibrated and validated on an extensive dataset, including error identification and refinement. The expected outcome is an optimized, efficient solution that minimizes time delays and additional costs in the creation of digital road environment models.本研究目标聚焦于一款用于自动检测选定交通基础设施要素的工具的开发与测试，其中包括测定其空间坐标与高程坐标。核心成果将是一款面向算法的测试接口，该算法旨在自动化处理用于构建数字环境模型的异构数据集，相较人工处理，可大幅节省时间与成本，同时提升模型质量与精度。所生成的模型可应用于交通模拟器及其他使用虚拟地形模型的场景中。 本研究的技术路径基于通过航空摄影测量与地面三维扫描获取的高分辨率栅格正射影像（raster orthophotos），同时测试替代成像技术与低质量输入数据，以评估系统的鲁棒性与相关风险。 在方法论层面，该系统将识别代表选定道路基础设施要素的特定像素色彩模式，在图像中对其进行聚类与识别，并将栅格数据转换为矢量（vector）格式。其创新点在于，仅对包含关键目标的选定图像区域执行自动矢量化操作，而非对整个栅格图像进行处理。该算法将在包含误差识别与优化环节的大规模数据集上进行校准与验证。 预期成果为一款优化后的高效解决方案，可最大限度降低数字道路环境模型构建过程中的时间延误与额外成本。