Expressway Data De-noised Approach Based on Wavelet Analysis

摘要

Compared with traditional collection methods using detectors or sensors, floating-car data collection has advantages such as a broad coverage area, low cost installation and maintenance costs and little influence from external sources. Therefore, floating-car data is being broadly applied in traffic relative studies. Traffic original data usually contains some random components, which disturb the distinction of traffic status. The original data must be de-noised in order to filter these random components. This paper presents wavelet analysis to de-noise the original data from floating-car data collection. In this paper, the data from RTMS (Remote Traffic Microwave Sensor) collection is treated as true data. Traffic original data, from various ring mainlines of Beijing on various days, are processed through wavelet threshold de-noising. The experiment results show that de-noised data is better than the original data in terms of evaluation indexes. For example, MSE of the de-noised data can drop as low as 6.36% and the correlation coefficient can be increased by more than 1.18%.