Ines Fenni

Parallelized multilevel Characteristic Basis Function method applied to scattering model for forest remote sensing

A parallelized version of multilevel Characteristic Basis Function method (ML-CBFM) is applied in this paper to the problem of electromagnetic scattering in the VHF and UHF bands from trees in large forest areas that are modeled as three dimensional, finite-length, dielectric cylinders. The use of this method enables us to achieve a significant reduction in terms of computing time and memory consumption, as compared to the conventional Method of Moments (MoM), and this, in turn, enables us to handle very large forest areas, well beyond the reach of conventional MoM. Furthermore, we take advantage of the fact that the ML-CBFM algorithm is naturally parallelizable, which provides us an added advantage over the conventional MoM. The ML-CBFM results are shown to be in good agreement with those obtained via the conventional MoM, which confirms the fact that the proposed method is not only computationally efficient, but is accurate as well.

Efficiency Enhancement of the Characteristic Basis Function Method for Modeling Forest Scattering Using the Adaptive Cross Approximation Algorithm

This communication discusses the hybridization of the extended version of the characteristic basis function method (CBFM-E) with the adaptive cross approximation (ACA) algorithm in the context of 3-D modeling of the problem of scattering from a forest environment. The ACA is applied when generating the reduced matrix to improve the CPU time associated with this step. The performance enhancement of the CBFM solution resulting from this hybridization is evaluated, and the impact of the geometry and heterogeneity of a natural forest scene on the gain achieved via the use of the ACA algorithm and on the accuracy of the solution is studied. We show that the hybrid CBFM-E/ACA approach enables us to significantly reduce the CPU time needed to compute the reduced matrix Zc without compromising the accuracy of the solution. Furthermore, the efficiency of the enhancement technique is not affected either by the dielectric contrasts of the scatterers or by the nonuniformity of the mesh that it is used to account for the heterogeneity of the forest scene.

Fast parallel implementation for electromagnetic modeling of scattering from forest environment

A combination of the Characteristic Basis Function Method (CBFM) and Method of Moments (MoM) can reduce and solve à large-scale electromagnetic scattering problems. The 3D full-wave model for electromagnetic scattering from forest environment, based on the volumetric integral Equation formulation of the electric field, applied to CBFM and implemented in Message Passing Interface (MPI) parallelization is described in this paper.

Efficient computation of macro-domain basis functions when applying the characteristic basis function method to the modeling of forest scattering

In this work, we discuss the use of diagonal representation (DR) of the macro-domain basis functions, in the context of the characteristic basis function method (CBFM) applied to forest scattering problem. We examine the suitability of this enhancement technique for heterogeneous natural forests. Results show that the use of the DR for the MBFs, according to certain criteria, enables us to significantly reduce the CPU time needed to generate the CBFs with little loss of accuracy, and without increasing memory costs. Consequently, the strategy of using the DR for the computing of the CBFs enables us to simulate larger forest scenes and to do so at higher frequencies.

A high performance MPI implementation of numerical modeling of electromagnetic scattering from forest environment

The Message Passing Interface implementation ensures a reduction of the computing time and a distribution of the necessary memory to each processor. The 3D full-wave model for electromagnetic scattering from forest environment, based on the volumetric integral Equation formulation of the electric field, combined with the Characteristic Basis Function Method (CBFM) and implemented in Message Passing Interface (MPI) parallelization is described in this paper. This combination of the CBFM and the Method of Moments (MoM) allows us to reduce significantly the size of the initial electromagnetic scattering problem.

Fast computation of macro-basis functions in the context of modeling the problem of scattering from forests

The aim of this research work is to discuss the use of the diagonal representation (DR) for the construction of the macro-basis functions (MBFs) when applying the Characteristic Basis Function Method (CBFM) to the problem of modeling of scattering from forests. We have applied the CBFM to improve the performances of a previously studied 3D full-wave model in order to solve electrically large forest simulation scenes. This efficient domain decomposition method has shown good performance both in terms of CPU time and memory while achieving a satisfactory level of accuracy compared to that of a conventional Method of Moments (MoM). Therefore, it enables us to solve forest EM problems of upward of 2 million unknowns by consuming only a reasonable amount of CPU time and using just 48 GB of shared memory (Fenni I.; Roussel H.; Darces M.; Mittra R., AP IEEE Trans., 62–8, 4282–4291, 2014).

Fast computing of large 3D dielectric forest scattering problems using the Characteristic Basis Function Method with the Adaptative Cross Approximation algorithm

This paper presents the hybridation of the Characteristic Basis Function Method with the Adaptative Cross Approximation algorithm while generating the reduced linear equation system. The proposed method is applied to 3D scattering model in the context of forest remote sensing. It enables us to realize a significant improvement of the performances of the CBFM both in terms of memory use and CPU time while maintaining a good level of accuracy compared to the CBFM solution.