Map Matching Model Based On Urban Floating Car Data

GPS trajectory data is a moving position point with a time stamp collected by a GPS recorder or GPS receiver.It has the characteristics of simple collection and low cost.Because of the positioning error in the satellite navigation system,there is a certain distance between the GPS track points and the reallocation points,which can not be directly applied to the construction of an intelligent transportation system.These GPS track points need to be corrected to the correct location.It is mainly done utilizing hardware and mathematical calculation,but the hardware correction method is expensive and challenging to improve correction accuracy significantly.Therefore,most research focuses on correcting the trajectory points using some mathematical methods,and the resulting GPS trajectory point correction method is called the map-matching algorithm.The general map matching method considers the spatial positioning accuracy of GPS track points,judge the distance between track points and adjacent road sections,and selects the nearest road section as the matching road section among all the adjacent road sections.Although this method can realize fast track matching,the consideration factor is single,and the result after track matching is always not satisfactory.Based on this method,many advanced map matching algorithms have been proposed.However,most map matching methods can not balance the matching efficiency and accuracy effectively.The matching accuracy is poor when dealing with GPS tracks with a low sampling rate.Given the above problems in map matching,this paper studies from the following points:(1)preprocessing GPS track raw data and urban road network raw data,eliminating track redundant points and noise points,sorting track points according to time sequence,Analyzing the Spatio-temporal characteristics of the road network,and building the topological relationship between road network nodes and road sections.(2)To improve the algorithm’s performance,the Douglas-Peucker algorithm dilutes the original trajectory points.The thinned trajectory can retain the essential space-time features of the original trajectory.Since the original road grid division method can not obtain the candidate road sections by track points,this paper proposes to use the Geohash coding method to grid the road network,each grid has a unique coding value,and the candidate road sections can be quickly found through the Geohash coding value when obtaining the candidate road section set.(3)Based on the traditional hidden Markov model,an HMM-CRFs hybrid map matching method is proposed,comprehensively considering the GPS trajectory’s Spatio-temporal context information.The problem of "label bias" when calculating the transition probability in HMM model is solved.When calculating the state transition of road sections,the weight adjustment coefficient is added,and the CRFs model estimates the parameters in the model.In addition,the driver’s travel preference is considered in calculating the state transition of road sections.A feature function,namely the shortest path distance between two candidate points,is added.Finally,the Viterbi algorithm calculates the maximum probability value of the global path,and the optimal matching path is found to realize the vehicle GPS trajectory.Two public data sets are used to verify the model.The results show that the method proposed in this paper can effectively balance the relationship between map matching accuracy and map matching efficiency.The correct matching percentage of sampling points is above 95%,and the average calculation time of each candidate road section is about 50 ms.While ensuring the efficiency of map matching to achieve the expected results,the accuracy of map matching is improved,and good matching can be achieved for complex intersections and parallel roads.The research results of this paper have particular practical significance in urban floating vehicle trajectory data mining and vehicle trajectory matching.