The Study On Genetic Algorithm Applied In Two-dimensional Inhomogenous Media Inverse Scattering

Electromagnetic inverse scattering aims to reconstruct the distribution of the dielectric characteristics of unknown inhomogeneous dielectric object by scattering wave. In the near thirty years, it has been extensively used in military, medical imaging and nondestructive testing field.The electromagnetic inverse scattering is one of most difficult problem due partly to its ill-posed , nonlinear and instablity nature. Traditional method can be classified two class: linear optimization technique and nonlinear optimization technique, linear optimization technique base on Born approximation or Rytov approximation is usually used to solve weak scattering problem. Nonlinear optimization technique can solve the strong contrast problem,wheares it spends much more time. In my article , GA is introdued to solve inverse problem.In my artice, positive scattering problem is solved by FDTD method and inverse scattering problem is sovled by genetic algorithm(GA)GA is a kind of searching method which simulates the normal evolution.In the solution of inverse problem convergence speed is significent and parameter setting has much effect on it. In present, no explicit regulation of parameter setting has been established. In article, basic regulation of the parameter setting is analyzed and some experience on it is obtained. Two method is propoesed in article to accerlerate convergence speed. The first method is calculating the individual's fitness again after crossover and decideing each individual's mutation range by the new fitness. After mutation, I check if the individual mutated is better? if it is better, perserved it as the next generation's individual,otherwise,the individual not mutated is perserved. The second method is calculating the "distance" between two individual,and finding out the similar individual. The better individual is perserved and worst individual is taken out. The two method mentioned above can accelerate the convergence speed and keep the diversity of colony and shorten calculation time.The validty of the algorithm is validated by the programming. Theresult of the reconstruction has been shown in chapter 5. And in conclusion I analyze the problem existed in using GA to optimize the electromagnetic inverse scattering and put forward the next research direction.