Methods of Real-Time Data Screening and Reconstruction for Dynamic Traffic Abnormal Data

Traffic data are the prerequisite and foundation of management and control for traffic flow. Because of the absence and abnormity of field vehicle detector data, four approaches for data screening were designed under considering the influence of interval of sampling and the inherent relationship between traffic three parameters including preliminary screening, threshold screening, traffic theory screening and statistical quality control screening. Then four approaches for data reconstruction were also proposed based on time series, historical data, spatial location and spatio-temporal correlation. Lastly a flow chart was set up for of preprocessing traffic data. Using the field data on city expressway in Beijing, algorithm can effectively remove abnormal data and the missing data prediction accuracy is less than 10%. These methods had been proved accurately, rapidly and stability, and can be used in engineering project.