Vehicles Detection in Complex Urban Scenes Using Gaussian Mixture Model With FMCW Radar

Vehicles detection via frequency modulated continuous wave radar has attracted a growing interest in recent years. However, the estimation of the distance and velocity of vehicles poses challenges under the complex urban scenes, as the returned signal is contaminated by the swaying trees and the buildings or the pedestrians. In this paper, an efficient method for vehicles detection is proposed to reduce the interference, especially caused by the moving foliage. Gaussian mixture model is established for the returned signal in frequency domain to segment foreground targets from the clutter background. Then, the motional parameters are obtained from the foreground curve extracted by Hough transform. The proposed method has a significant improvement in increasing the accuracy of the parameter estimation of vehicles in complex urban scenes. In addition, field experiments are provided to showing an outranking performance in comparison with conventional methods without decreasing the detection rate.