Research On Target Vehicle Trajectory Recovery Algorithm Based On Taxi Trajectory Data

With the development of mobile technology and the popularization of remote sensing positioning equipment,location-based services are becoming more and more popular.A large number of location-based applications emerge,generating a large number of GPS data and location check-in data.More and more data can be used to record the route of moving objects in the form of trajectories.These trajectory data should be of great research value for various location-based services.Trajectory recovery is an important method to ensure the quality of trajectory.The existing studies are all based on the historical vehicle track data to carry out algorithm training and obtain a path with the maximum posterior probability as the output.This paper proposes a Short trajectory recovery algorithm model(Short trajectory recovery system hereinafter referred to as SRS)based on the implementation of encoders and decoders,which is an end-to-end trajectory recovery method.Long-term memory network(LSTM)was used as model element.The encoder encodes the input locus to get the coding vector with spatio-temporal dependence.The decoder decoded the encoding vector to obtain the complete vehicle travel path.This paper conducts simulation experiments in real vehicle track data sets.Compared with the shortest path algorithm(SDD),the shortest pass-time algorithm(FP)and the most popular path algorithm(MPR),SRS has the highest accuracy,less running time and higher efficiency.Because the short path restoration problem is aimed at the path sequence restoration problem which only contains the starting point and the end point,but there are also long path restoration problems with multiple trajectory points.We cannot combine the simple short path recovery algorithm model,because this ignores the global spatio-temporal constraints.Aiming at the Long path trajectory restoration problem,this paper proposes a Long path trajectory restoration algorithm model(Long trajectory recovery system hereinafter referred to as LRS)based on attention mechanism.The encoder and decoder model combined with the attention mechanism solves the phenomenon of information loss caused by the increase of track input length.A comparative experiment was conducted on the real Harbin taxi track data set in this paper.The experimental results showed that the accuracy of LRS for long path track recovery was significantly higher than that of SRS as the track input length increased.The website of an MTV framework is designed and implemented.The user submits the track sequence to be recovered with Baidu API and the system displays it to users.