High accuracy GPS trajectory data can provide a lot of information for geographic information research and traffic analysis or other trajectory based work. However, in real life, the actual GPS data lose accuracy due to a lot of reasons such as low sampling rate, dense road network and abnormal GPS points. In this paper, a comprehensive map matching algorithm combined with crowdsourcing techniques is proposed to better perform the matching result. Map matching algorithms were originally designed to integrate the trajectory data and real-world road network data to identify the road link on which the vehicle is travelling and the vehicle location on that link. The basic idea is to add human cognition into this system. The main contribution of this thesis is to add a crowdsourcing procedu...[ Read more ]

High accuracy GPS trajectory data can provide a lot of information for geographic information research and traffic analysis or other trajectory based work. However, in real life, the actual GPS data lose accuracy due to a lot of reasons such as low sampling rate, dense road network and abnormal GPS points. In this paper, a comprehensive map matching algorithm combined with crowdsourcing techniques is proposed to better perform the matching result. Map matching algorithms were originally designed to integrate the trajectory data and real-world road network data to identify the road link on which the vehicle is travelling and the vehicle location on that link. The basic idea is to add human cognition into this system. The main contribution of this thesis is to add a crowdsourcing procedure with a voting strategy to the existing map matching algorithm. Evaluation of the algorithm with real data set illustrates that after the human annotation, the matching results are more reasonable, especially among the area with dense road network. For the future work, the algorithm can be incorporated with MTR crowd flow analysis, even supply background knowledge for other spatial crowdsourcing research.