Research And Application On Extraction Algorithm Of Urban Road Network Based On Vehicle Trajectory Data

Road network is an essential constituent part of a city's traffic system,and is the indispensable infrastructure for life of city's residents.Facing urban problems such as rapidly upgrading road network,traffic congestion,etc.,it has become increasingly a valued research topic as to how to update geographical data rapidly and accurately as well as how to preserve real-timeliness and accuracy of road information.The trajectories of vehicles can directly reflect the characteristics of road network,and track data has the advantages of being easily acquirable,low acquisition cost,wide road coverage,real time,etc.In the era of the big data era,compared to road extraction by conventional approaches of surveying and remote sensing image,the city's road network extraction method,which takes the vehicle trajectory data as the data source and technically backed by spatial data mining,has become increasingly a new idea for city's road network extracting and updating.Currently,road network extraction research based on trajectory data is gaining intensity.In terms of data,taxi data,due to its big data size,broad coverage,etc.,is widely adopted as experimental data.In this context,taxi trajectory data of Nanjing City is selected as object research data in this article.As to research method,in order to obtain relatively perfect road intersection data and effectively extract urban road network,the method of clustering algorithm to obtain intersection is used to construct road network.This article proposes and implements the city's road network extraction algorithm: “Data Pre-processing-Road Crossing Extraction-Road Centerline Extraction-Road Network Generation”.The nucleus of the essay research is embodied in the following three aspects:(1)Combined with the relevant tools such as visualization,etc.to analyze the characteristics of trajectory data.On this basis,further processing of trajectory data,mainly including: use threshold method to identify and eliminate drift data;DBSCAN clustering algorithm was used to identify redundant data areas and extract redundant key points for redundant data processing.(2)Acquisition of reliable information of road crossings is the key contents of road network extraction.Responding to the status of diverse traffic flows at arterial road,secondary trunk road and branch road of city's road network,this article sets forth DBSCAN Clustering Algorithm based on multi-density region division to realize the intersection of data clustering,using weighted centroid algorithm can obtain characterization of road intersection position of feature points,realizing the effective extractions of road crossings.(3)Using the trajectory points falling between road crossings as data source,this article sets forth two methods of obtaining centerline.Utilize buffer to extract skeleton line method to extract the center line of the road;Utilize Multivariate Adaptive Regression Splines(MARS)Method to fit road trajectory data to obtain road centerline.Meanwhile,have experimental and analysis on the two methods,so as to seek the best way to acquire road centerline.(4)Based on the identification of road intersections and the extraction of the central line of road sections,the relationship between intersections and road sections is studied,and the road network topology is constructed to form a city road network vector map.