Public Transit Network Generation And Map Matching Based On GPS Data

With the maturity of satellite positioning technology and the intelligent development of buses,GPS data collected by bus-borne equipment not only contains vehicle trajectory information,and these data have the characteristics of high accuracy and low collection costs.Generating public transit network based on massive bus GPS data is of great significance for adjusting the structure of the public transit network,rebuilding urban roads,alleviating traffic congestion,and realizing intelligent planning of bus lines,and it has always been one of the research hotspots in the field of intelligent transportation.In the research of generating a line network based on the GPS data of the bus,this paper focuses on the filtering of noise points in the GPS data,the parameter selection of the noise point processing algorithm,and matching the generated public transit network with the actual road.The main research work and innovations are summarized as follows:(1)Aiming at the problem of large amount of GPS data,uneven data density,and too much noise points to generate bus routes,a pixel-based fast density clustering(Pixel?Based-DBSCAN,PB-DBSCAN)algorithm is proposed.PB-DBSCAN represents the GPS data set as a set of two-dimensional pixel grids and changes the relationship between the judgment data points and other data points during the clustering process to determine the relationship between the current pixel grid and adjacent pixel grids,which greatly reduces The search range of GPS data points in the data neighborhood.PB-DBSCAN can effectively identify and filter the noise points in the GPS data set,reduce the time complexity of clustering and achieve fast clustering.(2)In the process of clustering GPS data of buses on different lines,the problem of large deviations in clustering results and poor clustering results due to the non-university of the same fixed clustering parameter on different lines,a method for dynamic parameter selection is proposed.This method introduces two definitions of clustering rate and optimal clustering rate interval.Through exploratory experiments on multiple routes,the optimal clustering rate interval can be obtained when it meets the requirements of the clustering effect of most routes.When the clustering rate is not in the optimal clustering rate range,the clustering parameters are dynamically increased or decreased according to the size of the clustering rate,so that the parameters are adaptive to different lines.Compared with the clustering algorithm with fixed parameters,the dynamic selection of parameters significantly improves the accuracy of clustering,the F₁ value increases by 10%on average,and the RMSSTD decreases by 30%on average.(3)Aiming at the problem that the bus lines generated based on GPS data do not match the actual roads,a map matching algorithm based on road scores is proposed.On the basis of the traditional geometric map matching algorithm,multiple factors such as the vehicle's driving direction and topological structure are comprehensively considered,which not only inherits the simple and fast characteristics of the geometric method,but also improves the accuracy of the algorithm,the accuracy rate increases by 10%on average.The map matching system was designed and implemented.First,the road network data was obtained from the OSM.To facilitate subsequent calculation and use,Osmapi was used to analyze the obtained OSM road network data.Second,the road network data was divided using Geohash.Finally,the road network data was imported into In the Postgre SQL database,and the functionality and algorithm accuracy and efficiency of the map matching system are verified and analyzed.