Flip procedure in geometric approximation of multiple-component shapes -- Application to multiple-inclusion detection

TL;DR

This paper introduces a novel geometric approximation method for multi-component shapes that adaptively detects topology changes and prevents boundary collisions, improving inverse obstacle detection accuracy.

Contribution

It proposes a flip procedure and intersecting control polygons detection within a shape optimization framework to better approximate shapes with unknown component counts.

Findings

Effective detection of multiple shape components

Prevents boundary double points during deformation

Accurate shape and topology reconstruction in inverse problems

Abstract

We are interested in geometric approximation by parameterization of two-dimensional multiple-component shapes, in particular when the number of components is a priori unknown. Starting a standard method based on successive shape deformations with a one-component initial shape in order to approximate a multiple-component target shape usually leads the deformation flow to make the boundary evolve until it surrounds all the components of the target shape. This classical phenomenon tends to create double points on the boundary of the approximated shape. In order to improve the approximation of multiple-component shapes (without any knowledge on the number of components in advance), we use in this paper a piecewise B\'ezier parameterization and we consider two procedures called intersecting control polygons detection and flip procedure. The first one allows to prevent potential collisions between two parts of the boundary of the approximated shape, and the second one permits to change its topology by dividing a one-component shape into a two-component shape. For an experimental purpose, we include these two processes in a basic geometrical shape optimization algorithm and test it on the classical inverse obstacle problem. This new approach allows to obtain a numerical approximation of the unknown inclusion, detecting both the topology (i.e. the number of connected components) and the shape of the obstacle. Several numerical simulations are performed.