Geometry Reconstruction of Conducting Cylinders Using Genetic Programming

A genetic programming-based method for the imaging of two-dimensional conductors is presented. Geometry is encoded in this scheme using a tree-shaped chromosome to represent the Boolean combination of convex polygons into an arbitrary two-dimensional geometry. The polygons themselves are encoded as the convex hull of variable-length lists of points that reside in the terminal nodes of the tree. A set of genetic operators is defined for efficiently solving the inverse scattering problem. Specifically, the encoding scheme allows for a standard genetic programming crossover operator, and several mutation operators are designed in consideration of the encoding scheme. Several results are presented that demonstrate the method on a number of different shapes