Smoothing of vehicular trajectories under heterogeneous traffic conditions to extract microscopic data

Abstract

Trajectory data collected using video image processing techniques are prone to noise. Trajectory data extracted using commercially available video image processing software (TRAZER) contains the noise associated with the false detection in addition to the white noise. This paper proposes a method based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) to smooth such trajectory data. In this approach, trajectory data are decomposed into a finite number of intrinsic modes and a unique residue is computed to obtain each mode. This monotonic residue gives the smoothed trajectory. The instantaneous speeds of the vehicles are then estimated using the method of continuous wavelet transforms, discrete wavelet transforms, and numerical differentiation. Internal consistency analyses show that the wavelet transforms methods are effective in reducing the noise amplification of the speed profile. It was also observed that the corrections applied on trajectory data have a significant effect on macroscopic traffic relations.