Automatic Vehicle Monitoring (AVM) systems are exploited by public transport companies to manage and control their fleet of vehicles. However, these systems are usually based on the background knowledge of the transport network which can change during the time and in some cases can be missing or erroneous. GPS data and other information captured by the vehicles during their work can be exploited to update the network knowledge. This paper presents a novel approach, namely DeCoClu (Density Consensus Clustering), that aims at mining the topology of a public transport network by means of a consensus clustering density-based approach. In particular, the method exploits static information from time series of positioning signals (i.e., GPS data) to infer geographical locations of stops by means of a consensus clustering strategy based on a new distance function. Moreover, the logical pathway of a route (i.e., stops sequence) is defined by an Hamiltonian cycle. Experiments performed on real-data collections provided by a public transport company demonstrate the effectiveness of the proposed approach.