Review of road selection methods for purpose of multiscale mapping

Cartographic generalization is the process of reducing map details in order to present information at a smaller scale, ensuring the map remains coherent, topologically correct, and legible (Stanislawski et al., 2014;Kraak et al., 2020).Traditionally, cartographers decided which details to keep based on their expertise, making generalization both an art and science (Pawlak, 1971).Road networks are essential on general geographic maps, therefore selecting roads to display at multiple scales is of the key importance.AI techniques, especially machine learning (ML) and deep learning (DL), can assist this process.These methods use decision trees, neural networks, and graph convolutional networks to select road networks by assessing the road importance.Although these models seem promising, they face challenges in processing geographic data, in particular they require a large amount of ground true (reference) data as well as efficient and plausible evaluation measures.This research focuses on the assessment of methods for generalizing road networks treating road selection as a primary step.In this review across the Web of Science and Scopus databases we have identified 44 relevant studies (from 742 initially found using search queries) that demonstrated the effectiveness of different road selection approaches.