Extraction And Refining Road Networks From Lowprecision GPS Trajectory Based On Road Horizontal Alignment Model

How to obtain large-scale and current road network data in a timely manner to realize the automatic update of the road network has always been an urgent problem to be solved in the construction of smart transportation and smart cities.Crowdsource trace data represented by vehicle-mounted or hand-held GPS motion has the advantages of large volume,rich information,low cost,as well as timely and efficient,providing a new pattern for road network data acquisition.The geometric position,network topology,and traffic status of the road can be quickly and intuitively extracted by GPS track data,which has become a research hotspot in surveying & mapping and remote sensing profession and intelligent transportation industry in recent years.Low-precision GPS trajectory data has the advantages of better real-time performance and low cost,but the sampling frequency is low and the accuracy is not high.Existing researches on road network extraction based on GPS trajectories are mostly based on the spatial distribution and geometric characteristics of the trajectory data points,without considering the inherent geometric attributes and topological relationships of the road route,resulting in poor matching between the road network and the actual road,so its geometric accuracy and topology accuracy is often not guaranteed.Based the issues above,this study takes taxi GPS track data as the research object,and proposes a method for extracting and optimizing low-precision GPS track network based on road plane linear model.First,a filtering method based on angle threshold and DBSCAN density clustering is proposed to eliminate GPS drift and redundant data.Then divide the study area into grids,and extract the initial road network based on the improved DBSCAN algorithm;Finally,the feasibility and effectiveness of the method are verified through qualitative analysis and quantitative analysis,and the research results have certain theoretical significance and application value.The main work of this study are as follows:(1)A data preprocessing method is constructed.That is,first analyze the characteristics of the trajectory data,and then propose a filtering method based on angle threshold and DBSCAN density clustering to eliminate GPS drift and redundant data according to the data characteristics,and complete the data preprocessing process.(2)A road network identification method based on DBSCAN improved algorithm is constructed.That is,the research area is gridded first,and then the number of track points in each grid is counted and the density difference between adjacent grids is calculated,select the grid with the highest density as the initial grid and merge the grids that meet the density difference to obtain a relatively uniform density area.Then repeat the above steps until you can no longer merge to obtain multiple grids with uniform density and gradually decreasing,calculate the clustering parameters of the grid with the largest density.Other grid parameters can be determined according to the density relationship between the grids to complete the data clustering.Finally,the angle and distance are used as constraints to gradually obtain the road network using a recursive method.(3)A road network optimization method based on road alignment model is proposed.That is,firstly,based on the principle of principal curve analysis,the decision function is introduced to segment the road to obtain the road segmentation map.Then a seven-point fitting circle curvature algorithm is proposed for line element recognition.Finally,the average value of the operating speed of each road segment is introduced,and each linear element is adjusted according to the road horizontal linear model to obtain the optimized road network data.Taking the taxi trajectory data of Wuhan for seven consecutive days as the research object,the optimization results are superimposed on the OSM map and the satellite image map for qualitative analysis.Take OSM map as the standard map and other road network extraction methods for quantitative analysis,the results signify that the road network data obtained by the algorithm in this paper has been significantly improved in geometric alignment and topology accuracy.