Review of road selection methods for the purpose of multiscale mapping

Cartographic generalization reduces map detail to ensure clarity and accuracy at smaller scales. This study reviews road selection methods for multiscale mapping, covering scales from 1:10,000 to 1:1,000,000. The applied methodology was inspired by the PRISMA search model and based on searches in the Web of Science and Scopus databases, supported by the ResearchRabbit platform. As a result, five categories of approaches: semantic-based, stroke-based, mesh-based, graph-based, and machine learning-based, were identified and analyzed in terms of their strengths and limitations. Emphasis was placed on automation and the importance of selecting appropriate generalization techniques based on map scale and purpose. The literature review also revealed a variety of quality evaluation metrics used in the analyzed approaches, with a predominance of quantitative measures such as accuracy and F1-score, complemented by qualitative expert visual assessments.