An Accurate and Fast Global Map-Matching Approach for Low-Sampling-Rate GPS Trajectories

Current global matching algorithms for low-sampling rate data still face the challenges of large computational costs and significant declines in accuracy with increases in the sampling interval time. Therefore, this paper proposes a novel global map-matching algorithm based on the hidden Markov model to match the data accurately and quickly. First, we utilize the road accessibility instead of the gridding or confidence intervals to obtain the candidate information. Then, this algorithm synthetically considers the trajectory location at the previous and next moments of the current trajectory point to establish the candidate paths. Finally, this algorithm calculates the transition probability through a similarity calculation between the spatial-temporal position of the candidate paths and the GPS trajectory. The accuracy, calculation time and stability of the proposed algorithm are verified systematically. For different sampling intervals, compared with the testing algorithms, the proposed algorithm improves the mean accuracy by a minimum of 3.3% and a maximum of 32.4% for different sampling intervals. The computational efficiency of the proposed algorithm is significantly improved by 3-391 times. It provides a new method for the development and analysis of low-sampling rate data.