Ground Clutter Detection Using a Gaussian Mixture Model

A novel 3-dimension (3D) discriminant function is introduced in this paper to improve ground clutter detection for weather radar observations. It is found that the phase variation of clutter signals is different from that of weather, and hence a phase fluctuation index is defined and used as a discriminant for clutter detection. The phase fluctuation index is combined with the dual-scan cross-correlation coefficient to form an exceptional 3D discriminant function to achieve a high probability of clutter detection. An optimal decision based on the proposed 3D discriminant function is made by the Bayesian classifier to identify clutter from weather signals. To make it more efficient to use, a multivariate Gaussian mixture model is used to parametrize the probability density functions (PDFs) of the discriminants for the clutter and weather signals. The model parameters are estimated based on the maximum likelihood method. The performance of the proposed algorithm is shown by applying it to the radar data collected by the polarimetric KOUN radar and is compared to that of several existing detection algorithms. The comparison results indicate the superiority of the proposed clutter detection algorithm.