Abstract
Microwave imaging of parallel perfectly conducting cylinders using scattering data with or without the effect of random noise is considered in this paper. A novel algorithm, the real-coded genetic algorithm coupled with Newton-Kantorivitch method (RGA-NKM) is proposed to deal with it with significantly improved performance. The main idea of the RGA-NKM is to perform a Newton-Kantorivitch type search for the local optimum after the genetic operations in each genetic evolution. Numerical results and comparisons with both real-coded genetic algorithm (RGA) and Newton-Kantorivitch method (NKM) demonstrate that although the simplicity of RGA is lost, the search ability is greatly improved and the convergence is sped up significantly while those merits of RGA such as high level of robustness, versatility and insensitiveness to ill-posedness are retained.