Microwave image reconstruction of 3-D dielectric scatterers via stochastic optimization approaches
- Publication Type:
- Issue Date:
The reconstruction of microwave images is generally considered as a nonlinear and ill-posed inverse scattering problem. Such problems are generally solved by the application of iterative numerical methods. However, the accuracy of images reconstructed by traditional methods is heavily dependent on the choice of the initial estimate used to solve the problem. Thus, with the aim to overcome this problem, this research work has reformulated inverse problems into global optimization problems and investigated the feasibility of solving such problems via the use of stochastic optimization techniques. A number of global inverse solvers have been implemented using different evolutionary strategies, namely the rivalry and cooperation strategies, and tested against a set of imaging problems involving 3-D lossless and lossy scatterers and different problem dimensions. Our simulation results have shown that the particle swarm optimization (PSO) technique is more effective for solving inverse problems than techniques such as the genetic algorithms (GA) and micro-genetic algorithms (μGA). In addition, we have investigated the impact of using different PSO neighborhood topologies and proposed a simple hybrid boundary condition to improve the robustness and consistency of the PSO technique. Furthermore, by examining the advantages and limitations of each optimization technique, we have proposed a novel optimization technique called the micro-particle swarm optimizer (μPSO). With the proposed μPSO, excellent optimization performances can be obtained especially for solving high dimensional optimization problems. In addition, the proposed μPSO requires only a small population size to outperform the standard PSO that uses a larger population size. Our simulation results have also shown that the μPSO can offer a very competitive performance for solving high dimensional microwave image reconstruction problems.
Please use this identifier to cite or link to this item: