Abstract
We propose an algorithm that can be used to identify automatically the subset of street segments of a road network map that corresponds to a junction. The main idea is to use turn-compliant locomotion affordances, i.e., restricted patterns of supported move- ment, in order to specify junctions independently of their data representation, and in order to motivate tractableindividuation and classification strategies. We argue that common ap- proaches based solely on geometry or topology of the street segment graph are useful but insufficient proxies. They miss certain turn restrictions essential to junctions. From a com- putational viewpoint, the main challenge of affordance-based individuation of junctions lies in its complex recursive definition. In this paper, we show how Open Street Map data can be interpreted into locomotion affordances, and how the recursive junction definition can be translated into a deterministic algorithm. We evaluate this algorithm by applying it to small map excerpts in order to delineate the contained junctions.
Highlights
Introduction and motivationRoad network maps, consisting of vector representations of roads, junctions, and points of interest (POIs), are among the most widely used sorts of geodata
We argued that current approaches ignore a crucial source of information (Section 2), namely information about turn-compliant locomotion affordances
The channel network theory is a semantic theory that directly represents these affordances [18]. We showed how this theory can be interpreted into a road network database like Open Street Map (OSM), and how a recursive definition of an n-way junction can be translated into an algorithm with testable correct results, and with acceptable runtime behavior for small map excerpts
Summary
Road network maps, consisting of vector representations of roads, junctions, and points of interest (POIs), are among the most widely used sorts of geodata. Navigation systems need detailed representations of highway intersections in order to guide drivers through the junction’s complicated sub-network All this requires identification routines for junctions at different levels of detail. On a high level of generalization, category instances may be represented as single vector data elements (e.g., single points for junctions, or single segments for roads). In order to approach this problem independently of a certain kind of data representation, road networks can be considered in terms of data semantics [17]. This definition is not based on segment geometry or topology, but on observed affordances It is independent of a specific level of representation, while capturing the intended meaning in terms of possible navigation manoeuvre.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
