Blind Information Processing Aigorithms And Applicationin Traffic Network Based On Massive Spatio-temporal Data

The intelligent transportation system can improve traffic efficiency,it is an effective method to solve the bottleneck problem of traffic congestion in the city development. But the normal operation of the intelligent transportation system must rely on the electronic map which can describe traffic network structure accurately and real-timely.When constructing urban traffic network map in traditional methods,fuse the remote sensing data and probe vehicle data by map matching model, it requires a lot of time to collect data and two sources of data may cause errors. The traditional map construction method can not reflect the morphological changes of traffic network in time, and the map data is not accurate enough.The prerequisites of updating urban traffic network map data timely include two aspects, that is to get a single vector road (such as road sections) structure data and the core points (such as traffic lights) position data. Floating car track location data can describe the structure of the road. In addition, the contains the GNSS (Global Navigation Satellite System, global navigation satellite system) transient velocity, transient direction angle and other driving characteristic data can also describe the floating car driving state. Some information parameters of the road can be estimated according to the driving state of the floating car on the road. Above all, the floating car track location data is the ideal data source to construct and update traffic network map.This paper propose an algorithm based on spatio-temporal data to process traffic network blind information. Propose blind information processing theory firstly, analyzes the relationship between spatial position distributions of floating car data and the meeting relationship between floating car and vector road in space,and then design a method for separating the abnormal traffic data by Time-segment separating, obtain the observed data without obvious abnormality. Propose the neighborhood centroid clustering method by experiment and get vector road structure core points by linear interpolation and arc interpolation. Finally, according to the analysis of driving characteristics of bus and taxi at core point position, put forward the method of extract road core points by neighborhood average speed, and the extraction results are modified according to the symmetric feature of the vector path.The vector road structure and key points position of road (Bus stops position,traffic lights position)are extracted by data mining method of blind information without manual measurement data in this paper. Experiments were carried out in combination with the Jinan taxi and bus GPS data. The results show that the method to construct the vector road with high accuracy, can accurately describe the real road feature,the average error of the distance from the road bus stops to real position is 6.31 meters, and the average error of distance from traffic lights to real position is 7.58 meters,data update frequency is 10 days/times,it meets the requirements of fast and accurate construction and iterative updating of traffic network.