| Since economic reform and open up,great changes have taken place in China’s socio-economic and industrialization process,and the amount of mobile pollution sources has increased dramatically,resulting in a large number of air pollutants,which have a serious negative impact on human health and the environment.It is urgent to monitor and control them effectively.Mobile Pollution Source Monitoring Station has positioning equipment,so it is very convenient to obtain spatiotemporal trajectory data.Spatiotemporal trajectory data contains three attributes:moving object,location and time.Time-space trajectory data in urban computing and location prediction of moving objects have been widely used.Through the study of GPS time-space trajectory sequence data of mobile pollution sources acquired by monitoring stations,the trajectory of mobile pollution sources can be tracked and matched in real time,and the short-term trajectory of mobile pollution sources can be predicted,thus providing effective monitoring technology for relevant departments.This paper analyses the factors that affect the matching accuracy of the tracking and matching algorithm of mobile pollution sources,and studies and summarizes the existing algorithms.Aiming at the problem of mismatch between the trajectory of mobile pollution sources and the actual road network caused by GPS sampling error and relative position error,etc.This paper presents a real-time trajectory tracking matching algorithm based on network topology and weight.First,the topological structure of road network is constructed using breadth-first traversal algorithn,and then the approximate set of road segments to track the matching trajectory is selected by using topological constraints and spatial constraints.Then,the distance weight,direction weight and relative position relation weight of each section to be selected are added together,and the total weight value is taken as the weight condition for solving the shortest path.Finally,Dijkstra algorithm is used to calculate the optimal tracking matching segment sequence.Compared with the traditional two algorithms,the proposed algorithm has higher accuracy and shorter time-consuming under the same data source through simulation experiments.Furthermore,aiming at the problem of accurate and real-time trajectory prediction of mobile pollution sources,a hybrid intelligent genetic particle swarm optimization algorithm is proposed based on the research and summary of existing trajectory prediction algorithms.Traditional Extreme Learning Machine(ELM)has the problem of poor generalization performance of small data sets.Although Optimized Extreme Learning Machine(OELM)overcomes this shortcoming,it of ten fails to achieve optimal results in the prediction process due to the influence of input weights and the randomness of hidden layer node bias assignment.Therefore,Hybrid Genetic Particle Swarm Optimization(HGPSO)is used to dynamically optimize the parameters of OELM model(input weights and hidden layer node bias).To improve the randomness of the model,the mirnber of hidden layer neurons needed by the model is fewer,and the generalization of the network is improved.Compared with the existing five algorithms through Simulation experiments,the algorithm has better prediction accuracy and real-time performance,and on this basis,multi-step time series prediction is realized.Finally,the tracking and matching algorithm and the prediction algorithm of the trajectory of the mobile pollution source are engineered into the online monitoring system of the mobile pollution source.The algorithm runs in the production environment for a long time,showing good real-time and stability. |
