Underground Road Detection Based On Trajectory Data

The development of satellite positioning technology and the popularity of location-aware devices have promoted the rapid generation of massive trajectory data.By ana-lyzing and mining the trajectory data information,we can detect the users,hobbies and behavior habits,master the trend of urban population migration,and monitor urban traf-fic conditions in real time,etc.Although the city electronic map has been widely used,users often encounter navigation information error due to the rapid development of munic-ipal construction and the update lag of electronic map.Especially in underground roads,GPS devices cannot navigate normally due to the shelter of ground,resulting in frequent occurrence of traffic accidents.Therefore,it's an urgent problem to be solved to detect underground roads without delay and estimate objects' locations accurately on the roads which are prone to lose GPS signals.This thesis detects underground roads based on the trajectory data.Because there are information errors in the trajectory data,this thesis firstly designs trajectory feature rules based on the Map-Reduce distributed framework to preprocess the trajectory data.Then,we propose two kinds of underground road detecting methods for having electronic map and no electronic map respectively.Finally,the results of the underground road detecting methods and the existing location prediction algorithm are combined to solve the problem that moving objects can't be positioned on the roads which are prone to lose GPS signals.Main contributions of this thesis are as follows:· Underpass detecting algorithm with electronic map Because the city electronic map information is not complete and is updated with delay,the information of some underpasses may be missing.This thesis proposes an underpass detecting algorithm based on the trajectory data and the electronic map,which can find underpasses that haven't been marked on the map.Experiments demonstrate the effectiveness and efficiency of this algorithm.· Underground road detecting algorithm without electronic map This thesis pro-poses an underground road detecting algorithm without road network,which can discover underpasses for pedestrians and tunnels for vehicles.And an R-tree index is built to optimize the efficiency of the algorithm.· Actual location estimating algorithm When the satellite navigation signals are unstable or missing,it's very important to estimate the actual location.This the-sis proposes a multi-step location estimating algorithm.It firstly improves the ST-Matching map matching algorithm to adapt to the distributed computing cluster.Then it combines the GPS signal losing probabilistic distribution with the existing method to solve the problem that moving objects can't be positioned in the under-ground roads and other roads which are prone to lose GPS signals.