基于RELAX和PSO算法的GTD模型参数估计

Parameter estimation of GTD model combining RELAX and PSO

针对传统的几何绕射理论(geometric theory of diffraction,GTD)模型参数估计方法存在模型定阶困难、低分辨率时多散射中心难以准确估计、计算量大、易收敛于局部极值等问题,提出一种组合松弛(RELAX)算法和改进粒子群算法(particle swarm optimization,PSO)的GTD模型参数估计新方法。该算法基于RELAX思想,可以有效解决模型定阶问题;通过成像处理给定距离参数初值,并引入变异机制,使之能够稳定高效地收敛于全局最优值;在每次估计时通过检验同一分辨单元内是否存在多个散射中心,使之具有较好的超分辨能力。实验结果表明,该算法可以高效准确地估计GTD模型的散射中心参数。

Aiming at the difficulties in determining the order of the geometric theory of diffraction（GTD） model,distinguishing multiple scattering centers in low resolution,huge computation and escaping from the local supreme value,a new method for parameter estimation of GTD model combining RELAX and improved particle swarm optimization（PSO） is proposed.The iterative strategy of RELAX is applied to avoid the order determination of model,and the algorithm stably and efficiently converges at the global optimum value by giving the initial value of range parameters through imaging process and introducing the mutation mechanism.The super resolution is achieved by examining whether one scattering center or more exists in a resolution unit.The feasibility and effectiveness of the proposed method are verified by simulation and measurement.