Improving the effectiveness of GA-based approaches to microwave imaging through an innovative parabolic crossover

Several studies have shown that evolutionary-based approaches are efficient, effective, and are robust optimization methods for microwave imaging. However, the convergence rate of such techniques still does not meet all the requirements for online real applications and attempting to speed up the optimization is needed. In this paper, a new local search operator, the fitness-based parabolic crossover, is proposed and embedded into a real-coded genetic algorithm. Such a modification enables the imaging method to achieve a better tradeoff between convergence rate and robustness to false solutions. By exploiting the relationship between the crossover operation and the local quadratic behavior of the functional, it is possible to increase the convergence rate of the genetic algorithm and, thereby, to obtain an acceptable solution with a smaller number of fitness function evaluations. The effectiveness of the modified genetic-algorithm-based imaging method is assessed by considering some synthetic test cases in different dimensions and noisy conditions. The obtained numerical results provide an empirical evidence of the efficiency and reliability of the proposed modified evolutionary algorithm.