ABSTRACT
ABSTRACT: When generalizing from large scale databases the problem of road junction generalization arises: the road junction representation needs to be simplified, removing all the road segments composing the junction except those representing the main roads. This paper will present an approach to detect and generalize road junctions developed inside the CARGEN project at the University of Padova (Italy) and suitable to generalize road junctions from 1:5000 scale to 1:25000 scale. The adopted approach detects road junctions starting from the analysis of cycles in the road graph. Then, applying a morphological analysis, we are able to classify road junctions in different types allowing ad-hoc generalization algorithms to be applied to each junction type. The paper will illustrate how this approach achieves the aimed results and at the same time avoides to generate topological errors.1
