Generation Of Road Network Information From Massive Less Frequent GPS Vehicle Trajectories

Coincident with the booming of the economy and development of society is the expansion of cities' transportation systems all over the country.Lots of new roads are constructed,while a large number of old roads are repaired and updated every day.Apparently,the city traffic network structures now change rapidly in China.Thus,it is of great importance to collect and generate road network information timely and accurately.Plenty of applications,such as intelligent transportation system and Urban planning system,are based on that vital information.At present,digital road information is usually collected by professional survey vehicles with highly precise yet expensive GPS equipments.There is no doubt that this method is costly in time and resources.Hardly can it provide users the latest road information on account of the time-consuming process.Recently,remote sensing technology is growing rapidly.Multitudes of researchers have presented some methods to extract road information from satellite remote sensing images automatically or semiautomatically.However,a satellite image always contains multiple land cover types.It is difficult to extract precise road features from the images because of other features,distraction.Besides,satellite images are of high cost,too.With the development of mobile positioning technology and communication technology,obtaining traces of moving objects is not a difficulty any more.Ordinary GPS equipments are cheap enough to buy for civilian vehicles.Some industries and companys may have collected massive GPS data of vehicles which contain a large amount of information.Obviously,the trajectories of vehicles can reflect the location and shape of roads.Therefore,we can design some methods to generate road network map from those trajectories,especially taxi trajectories since the data is always up to date,abundant as well as low cost.In this paper,I first analyze the current state of road extraction research.Then,the technology and theory base,including mathematical morphology and multivariate adaptive regression splines,is introduced.Next,a method to extract road intersections based on mathematical morphology is proposed,in which the key innovation point is the extraction and result fusion from multiresolution trajectory iamges.Afterwards,two methods of generating road center line are raised and compared.Finally,I design a total framework of generating city road network from low frequency and low accuracy GPS trajectories.