Research On Road Network Extraction Based On Walking GPS Trajectory

Intelligent transportation system is beneficial to improve the urban trafficenvironment, reasonable distribution of transportation resource, which can obtaingreat economic and social benefits. Electronic map is the foundation of theconstruction of intelligent transportation system. Generating electronic mapautomatically and updating the information of road network timely have importantapplication value and significance for vehicle navigation, transportation planning andland using. Currently, most map generation methods based on remote sensing imageare only for some certain types of roads, that is, these methods are not very accuratedue to their universality, and the long image updating cycle makes the information ofroad network hard to update timely.With the rapid development of the intelligent mobile terminal and the mobileInternet, people can get GPS trajectory data more easily. GPS trajectory data has thecharacteristics of low cost, fast updating and wide coverage, etc. So, extractingmethods based on GPS trajectory attracted wide attention in academia and industry.Researchers often mine urban truck road network base on floating car or taxis’ GPStrajectory at present. However, existing methods ignored the automatic extraction ofpaths, which are plentiful and change frequently, to ensure its complete and accurateare very important in many application areas such as earthquake relief, areanavigation and rural tourism.Aiming at this problem, we proposed a new method of road network extractionbased on walking GPS trajectory in this paper. Our experimental data is a walkinginspection data set collected in Hunan University of Science and Technology, which isrecorded by security guards.The total number of track points is17,46million.Ourmethod contains three parts: data preprocessing, road centerline generation and roadnetwork accuracy evaluation. Firstly, we remove the raw GPS trajectory data outliersby data preprocessing to insure data accuracy. Secondly, we generate road centerlinesautomatically and make the road network vectorization with GPS trajectory. Finally,the method of this paper is evaluated from qualitative and quantitative methodrespectively, which based on the references from Baidu map and related information. In this paper, three methods are adopted to generate road centerlines successively,such as trajectory clustering, cluster center segmentation and centerline fitting. Theresults show that the proposed method not only is able to accurately extract the roadnetwork, coverage rate can reach96.21%while error detection rate is3.26%, but alsocan extract pathway and update road network.