Recognition Of Flooded Roads In The Situration Of Rainstorm By Trajectory Data

With the intensification of global climate change and the acceleration of urbanization,the frequency and intensity of extreme rainfall events have further increased.Urban waterlogging disasters caused by heavy rains often cause urban traffic interruptions,economic losses,and even casualties.Therefore,how to capture road traffic anomalies during the rainfall process and grasp the impact of heavy rainfall events on urban traffic and crowds is of great significance for reducing road traffic risks and ensuring urban operations.With the enrichment of multi-source perception data such as floating trajectory data and image monitoring data,road traffic research has a new perspective and data foundation.In view of this,this research combines multi-source data such as floating car trajectory to build an urban road traffic anomaly discovery model,deeply analyzes the urban road traffic anomaly law under heavy rain weather,and summarizes the geographical environment and road characteristics,and uses machine learning algorithms to explore similarities.The characteristic road realizes the discovery of the stagnant water road.The main research contents and results are as follows:(1)Based on massive multi-source data,we analyzed the data framework for the research on road traffic anomaly recognition of heavy rain,and realized the preprocessing of floating car trajectory data,social perception data such as Weibo,and meteorological and hydrological data such as rainfall and stagnant water.Invalid,error,redundancy and parking data in the trajectory data,the map matching of the trajectory data and the road vector data is completed,and the persistence of the trajectory data is realized.(2)Combined with floating car trajectory,rainfall data and Weibo data,the isolated forest anomaly detection algorithm is used to detect abnormalities in the road,and according to factors such as the time,frequency,and traffic characteristics of the abnormality,the road abnormalities are divided into regular types anomalies,accidental anomalies,and accidental rainfall anomalies.By comparing and analyzing the traffic characteristics of each anomaly,we concluded that the rainfall anomalies are caused by road water accumulation.(3)Baded on the disvcovery of abnormal road traffic,this paper summarize the geographical environment and road characteristics,select the most relevant features of the seizure road based on the random forest algorithm,use the GA-BP neural network to train the seizure road discovery model based on the selected most relevant features,and use the Weibo data to evaluate the model,and experiment The results show that the model has a recognition rate of 80.23% for the roads with stagnant water mentioned in Weibo.It effectively finds the stagnant water on the road under the rainstorm scenario,which helps to reduce the risk of urban travel under rainstorm.