Research On Map Matching Method Based On Track Segmentation And Improved HMM Model

The trajectory data is the spatial location points containing the information of speed and sampling time of moving objects collected by the equipment equipped with Global Navigation Satellite System(GNSS),which has the advantages of real-time and easy access.Due to the error of GNSS,if the original trajectory data is directly applied to the trajectory mining or intelligent transportation system,it will generate large errors,so it is necessary to use map-matching technology to correct the trajectory points to the real correct position of the vehicle.To address the problems of repeated matching of trajectory points and long traversal time of trajectory points in common map matching methods,as well as the poor effect of single feature trajectory segmentation method to deal with complex trajectories,this paper uses multiple spatiotemporal features of trajectory to segment the original trajectory and combines the improved hidden Markov model to realize map matching.The research results of this paper are as follows:(1)To ensure the accuracy and reliability of the input data for map matching.Pre-processing the original GNSS track data and the road network data of the study area,filtering noise points,redundant points,and reordering the track data;Coordinate system transformation and clipping are performed for road network data,and spatial index is established for road network R tree.(2)To reduce the impact of single-track points on map matching.A track segmentation method,CRHL,is proposed.This method considers the direction similarity and the length of the sub-track segment to limit the original track into several sub-tracks based on the position change rate,avoiding the problem of repeated calculation of track points.At the same time,the track segmentation also has a certain data cleaning function,which can reduce the subsequent matching calculation and improve the efficiency of the map matching algorithm.(3)Based on the traditional hidden Markov model map-matching algorithm,a map-matching method based on trajectory segmentation and HMM model is proposed.A series of sub-tracks divided by the original track is used as the observation sequence,and the candidate road sections corresponding to the sub-tracks are used as the hidden sequence.Calculate the similarity between the sub-track and the candidate road segment,and introduce the direction weight as the observation probability.When calculating the state transition probability,consider the accessibility between the candidate road segments of the two adjacent sub-tracks.Finally,the Viterbi algorithm is used to solve the joint probability maximum path as the best matching path.The experimental results show that the accuracy of the proposed map-matching method based on trajectory segmentation and HMM is more than 94%,and the average matching time of each candidate segment is about 39 ms.Compared with traditional HMM,this method has improved matching efficiency and accuracy and has a good matching effect for parallel sections,curves,and intersections.To sum up,the research methods in this paper have certain practical significance in map matching and trajectory data mining analysis.