A comparative study of two-dimensional modeling methods for electromagnetic scattering data

The aim of this research is to model two-dimensional data encountered in electromagnetic scattering problems using model-based parameter estimation techniques. Once a highly accurate model is constructed from a few samples, the model can then be used to interpolate between or extrapolate from the original samples at any desired point and any number of times, thus reducing the amount of data needed to be stored in memory or required to be measured. An added advantage is that the computations required to be carried out on the numerical samples can instead be carried out on the analytical model, which may reduce the computational complexity. It is intuitive that a higher number of terms in the model, increases the accuracy, but additionally it has the unwanted effect of increasing the computational complexity and memory requirement as well. An additional goal, therefore, is to solve the optimization problem of obtaining a model by maximizing the accuracy and minimizing the number of terms. Several modeling techniques are compared in this study, especially those based on matrix pencil methods. Some techniques for optimizing their performance have also been suggested. The pros and cons of each method are also discussed. It is shown that using the suggested techniques provides us with better models, but some pointers are also provided towards investigating more viable alternatives.