Object-based model verification by a genetic algorithm approach: Application in archeological targets

Abstract A new target-oriented parameterization scheme, named the object-based model, is suggested to represent man-made or natural targets as regular shapes embedded in a two-dimensional resistivity background. The numerical values of the target parameters (size, depth, location and resistivity) are estimated in three steps consisting of conventional regularized inversion, exclusion of anomalous regions and delineation of target bodies. The method produces sharp edges and sharp variation in intrinsic resistivity between the targets and background. The number of target objects is decided by the visual inspection of the 2D resistivity section derived from the application of a conventional cell-based regularized inversion. The 2D background is also extracted from the same section. A genetic algorithm approach is used at the final stage to test a large number of distinct models. Each test model consists of the same number of objects buried in the 2D background. The size, depth, location and resistivity of the targets are estimated from a class of models generated by the application of biological rules. The derived images of buried bodies have sharp edges and can then be understood by engineers and archeologists. However, if the hypothesis about the ‘conceptual model’ is very different from the geometry of the subsurface, the proposed approach will not be able to produce satisfactory results.