Multiple Radar Subbands Fusion Algorithm Based on Support Vector Regression in Complex Noise Environment

Real noise environment may not be Gaussian due to the existence of man-made interferences, natural bursts, and so on. In this paper, a new multiple radar subbands fusion algorithm in complex noise environment is proposed. The considered complex noise includes impulsive noise and the mixture of impulsive and Gaussian noises. The proposed subband fusion algorithm consists of the following steps. First, the incoherent factors are estimated and the multiple subbands are compensated to be mutually coherent. Then, coherent subbands are fit by the geometrical theory of diffraction (GTD) model, where model parameters are estimated in the frame of sparse reconstruction. Different probability distribution functions are used to describe different noise environments, and the maximum a posteriori probability (MAP) estimate of scattering model parameters is achieved. It is shown that the MAP estimate can be obtained by iteratively solving a support vector regression problem. Finally, the fused full band is obtained by extrapolating the measured subband data based on the estimated GTD model. Validity and performance of the proposed algorithm are investigated by analytical data, simulated data, and static-range data.