08-26, 12:00–12:30 (Europe/Rome), Modulo 0
Connectivity of roads in a map is essential for many use cases including navigation. We present a graph-based solution to the road conflation problem which takes into account the connectivity of the road network. First, we generate a road network graph in both sources based on bifurcation points. Second, we carry out node and edge matching between the graphs where we follow shortest distance as a matching criterion. This is followed by the merging stage where graph edges with matching end nodes get conflated. Newly added roads are connected with the graph based on node and edge matching. We carry out experiments on conflating open source footway datasets from multiple cities with the OSM. The resulting conflated map contains up to 16x map feature improvements per city with geometrically accurate and smooth results around road junctions. Future work involves using different graph matching criteria to improve on the conflated output.
Esra Cansizoglu is a Machine Learning Engineer working in the Map Building group at Meta. Her area of expertise is in computer vision and 3D geometry. She holds a PhD in Electrical Engineering from Northeastern University and a MS in Computer Science from Boston University.
Maps and GIS Analyst at Meta