CellNet

Road networks are essential nowadays, especially for people travelling to large, unfamiliar cities. Moreover, cities are constantly growing and road networks need periodic updates to provide reliable information. We propose an automatic method to generate the road network using a GPS trajectory dataset. The method, called CellNet, works by first detecting the intersections (junctions) using a clustering-based technique and then creating the road segments in-between. We compare CellNet against conceptually different alternatives using Chicago and Joensuu datasets. The results show that CellNet provides better accuracy and is less sensitive to parameter setup. The size of the generated road network is only 25% of the networks produced by other methods. This implies that the network provided by CellNet has much less redundancy.