Research On Congestion Recognition Of Urban Roads Based On Storm Platform

At present,the serious traffic congestion has become a prominent problem faced by large and medium-sized cities,which hinders the orderly development of the new urbanization construction.With the increasingly serious traffic congestion situation in the city,how to effectively identify the traffic status in the region has become a hot spot in the research of intelligent transportation.The traditional identification scheme basically relies on fixed sensors such as induction coils and cameras to build a traffic state detection network.Due to the limitations of high construction and maintenance costs,such methods do not fully cover the entire urban road network.Floating car data has gradually become the most important traffic data source because of its low collection cost and wide coverage.Through the analysis of the trajectory data of the floating car,the traffic flow state information can be excavated to realize the accurate identification of the congestion road.Floating cars produce hundreds of millions of location information every day,and the number of floating cars is still increasing.Massive data poses a challenge to traditional data processing technology.How to improve the throughput and performance of the system has become a primary problem.The emergence of distributed computing provides a new solution for large data processing.This paper focuses on how to use taxi GPS data to identify traffic congestion,the main contents fall into as follows:(1)The original taxi GPS data is preprocessed,including data cleaning and map matching.The source and format of data are introduced,and the characteristics and cleaning process of invalid data and noise data affecting the quality of GPS data are analyzed.On the basis of studying several existing map matching algorithms,hidden Markov model(HMM)is applied to map matching of locus points,which corrections the location information of GPS trajectory points deviating from the road.(2)A road congestion identification algorithm based on SAGA-FCM is proposed.The average speed of the road is the main indicator of congestion identification,which is estimated by the average speed of each car.According to the fuzziness of traffic state,fuzzy C means(FCM)clustering algorithm is applied to fuzzy partition the traffic state.Because the FCM algorithm is sensitive to the initial values and easy to fall into the local optimum,the simulated annealing algorithm and genetic algorithm(SAGA)is combined to improve its shortcomings.Experiments show that the SAGA-FCM algorithm is more accurate than the FCM algorithm in identifying congestion status.(3)A congestion recognition system based on Storm was designed.For the problem of real-time processing of large-scale GPS data,the architecture of distributed stream processing system is discussed.The SAGA-FCM congestion recognition algorithm is implemented based on Storm,message middleware and spatial database.The WebGIS related technology is used to visualize the road network and congestion road on the web.