Finite-Difference Time-Domain Simulation of Scattering From Objects in Continuous Random Media

Transcription

1 178 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 40, NO. 1, JANUARY 2002 Finite-Difference Time-Domain Simulation of Scattering From Objects in Continuous Random Media C. D. Moss, Student Member, IEEE, F. L. Teixeira, Member, IEEE, Y. Eric Yang, Member, IEEE, and Jin Au Kong, Fellow, IEEE Abstract A three-dimensional (3-D) finite-difference time-domain (FDTD) scheme is introduced to model the scattering from objects in continuous random media. FDTD techniques have been previously applied to scattering from random rough surfaces and randomly placed objects in a homogeneous background, but little has been done to simulate continuous random media with embedded objects where volumetric scattering effects are important. In this work, Monte Carlo analysis is used in conjunction with FDTD to study the scattering from perfectly electrically conducting (PEC) objects embedded in continuous random media. The random medium models under consideration are chosen to be inhomogeneous soils with a spatially fluctuating random permittivities and prescribed correlation functions. The ability of frequency averaging techniques to discriminate objects in this scenarion is also briefly investigated. The simulation scheme described in this work can be adapted and used to help in interpreting the scattered field data from targets in random environments such as geophysical media, biological media, or atmospheric turbulence. Index Terms Buried random objects, FDTD methods, Monte Carlo methods, random media. I. INTRODUCTION NATURAL media such as the atmosphere, snow, vegetation, rocks, and soils are often inhomogeneous media that cannot be described in a deterministic manner, so a statistical model should be employed instead [1]. In analytical studies, statistical models (random medium models) describe a medium as an effective or mean permittivity (or permeability) with random fluctuations that are generated from a prescribed correlation function. A single realization (assuming spatial ergodicity) or ensemble of random media with correlation functions chosen to describe the media of interest are used to study the statistical properties of the scattered and transmitted fields. In these types of media, the inhomogeneities may be described in a statistical manner, similarly to random rough surfaces studies [2]. Manuscript received November 16, 2000; revised September 11, This work was supported by the Office of Naval Research under Grants N , N , DTRS D-00043, and DACA K C. D. Moss and J. A. Kong are with the Research Laboratory of Electronics and Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA USA. Y. E. Yang was with the Massachusetts Institute of Technology, Cambridge, MA USA. He is now with the Rannoch Corporation, Cambridge, MA USA. F. L. Teixeira is with the ElectroScience Laboratory and Department of Electrical Engineering, The Ohio State University, Columbus, OH USA. Publisher Item Identifier S (02) Understanding wave attenuation, scattering, and phase fluctuations introduced by these random media is important for, e.g., remote sensing system design [3] and characterization of communication links [4]. In particular, predictions of radar return from objects obscured by foliage [5] or buried under snow or grass [6] are clearly dependent on knowledge of the effects of such geophysical media on the scattering response. Synthetic aperture radar (SAR), for example, is a coherent process that constructs an image of a target based on phase and attenuation information. When the target is obscured by a random medium, the scattering characteristics of the medium (clutter) will deteriorate the quality of the reconstructed SAR image. Simulation of the phase fluctuation and attenuation characteristics of random media can be used to estimate the effect that the medium would have on radar response. In addition, the results could be used in principle to help the design of signal processing algorithms to compensate for the distortion and clutter caused by random media. In this work, a three-dimensional (3-D) finite-difference timedomain (FDTD) scheme [7], [8] is introduced to directly model the scattered field and radar cross section (RCS) of an object in a random medium where volumetric-scattering effects are important. The random medium parameters in this case will be approximately based on particular soil models, with subsurface sensing applications in mind. Soil naturally contains fluctuations in density, material, and moisture that may affect the scattering results. In addition, because the purpose of many applications is very often to detect objects buried by man, the soil between the target and the ground surface has usually been previously excavated and as such, it is not expected to have a stratified or homogeneous profile. Numerous studies have appeared in literature in recent years focusing on the modeling of subsurface problems, e.g., [9] [12]. Initial studies concentrated on approximate analytical techniques [13], whereas more recently, there has been a great deal of work on numerical methods such as the method of moments [14] and the FDTD method [15], [9], [16]. FDTD has been gaining popularity for these applications due to its ability to model complex geometries with relative ease. In these studies, the treatment of the background soil (host medium) has been evolving in complexity, from the simple homogeneous dielectric slabs [13] to lossy, dispersive media with discrete particles [17], [18] and random rough surfaces [19]. Recent work has also been done on an FDTD scheme [20] that models the soil around the target as a random medium, /02$ IEEE

2 MOSS et al.: FINITE-DIFFERENCE TIME-DOMAIN SIMULATION 179 Fig. 2. Direct layer reflection is suppressed in the results that follow. Fig. 1. below the surface. 500 MHz incidence ( = 301). Object is a 20 cm 2 20 cm 2 10 cm rectangular PEC, buried 0.6 m but restricted to discrete scatterers (random placement). A 2-D FDTD study of point scatterers in inhomogeneous (Gaussian) media has also been recently published [21]. In this work, our focus will be on the 3-D FDTD simulation of scattering from objects buried in continuous random media with spatially fluctuating permittivity. Experimental studies have shown that soil constitutive parameters are highly sensitive to moisture [22] and geophysical material [9], so it is prudent to consider the effects of soil parameter fluctuations. Note that from the application viewpoint of a subsurface scenario simulation, the results presented here are incomplete in the sense that we are intentionally not including effects of any interface roughness (and the associated surface-scattering clutter). This is done in order to isolate and better understand the effects of the random volumetric scattering itself. This is also because FDTD studies of random rough surfaces have been carried out before [23], [24]. To analyze a real world subsurface general scenario in full, however, it is clear that both effects should be considered and analyzed together. II. FDTD AND RANDOM MEDIUM MODELS We consider an FDTD computational domain as show in Fig. 1, containing a rectangular perfectly electrically conducting (PEC) object buried in a random medium half-space. The incident field is introduced on a Huygens surface A (see Fig. 1) enclosing both the target and the random medium half-space, using a total/scattered (T/S) field formulation [7]. To determine the incident field, we assume a TE or TM plane wave incident on the half-space, and then calculate it everywhere on the Huygens surface A using the half-space Green s function [12]. For the Green s function used on Huygens surface A, we incorporate numerical dispersion effects to minimize noise in the scattered field domain [25]. For incident field calculations, the permittivity of the half-space is taken to be the mean permit- tivity of the random medium. The scattered far field is also calculated in a similar fashion (again using the half-space Green s function) on the Huygens surface B in the scattered field domain [16]. The results of the FDTD method to be presented here do not include the direct reflection from the half-space interface (free-space/mean permittivity), as illustrated in Fig. 2. The interface return (surface scattering) is removed because we are only interested in the perturbations to the RCS response caused by the buried object and random medium. Hence, only the object return, volumetric random medium return (fluctuations), and object-random medium interactions (and their interactions with the half-space boundary) are included in the far-field results. If necessary, direct returns from the interface layer can be easily obtained (even analytically) and included in the simulations. The scattered fields shown here can therefore be considered as perturbations in the steady-state average return with no object present. To be consistent with the T/S field formulation (which uses a half-space Green s function), the random medium fluctuations are truncated in the FDTD simulation so that the incident field Huygens surface A encloses all the scatterers, including the random medium itself. This truncation introduces an approximation because the incident field is injected without taking into account the random fluctuations. In other words, at points where the incident field is injected at the T/S (Huygens A) surface surrounding the random soil, it is created as if it had propagated through a homogeneous medium. The truncation is expected to be more pronounced for larger incident angles (from normal), because the incident fields travel greater distances through the homogeneous portion of the half-space. This approximation is similar to a Born approximation, but less stringent since it is applied only to the Huygens surface A. This same applies for the scattered field directions, since the approximation is also used for the Huygens far-field transformation surface (Huygens surface B). The finite size of the random medium introduces further approximations, in close analogy to the classical truncation error effect observed in random rough surface simulations [3]. Larger domain sizes would scatter more power, as we have not normalized the scattered field with respect to the power incident on the volume. Compared to a larger domain, a finite size domain produces a smoothing of the scattered cross sections and therefore, scattering results should be interpreted with proper care. Nevertheless, because the main interest here lies on the interaction between a buried PEC object and the random medium, a plane wave excitation over a surrounding finite volume is a useful scenario. A more realistic simulation can use a tapered wave in-

3 180 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 40, NO. 1, JANUARY 2002 cidence, but this would greatly increase our computational demands because the plane wave allows a dimensional reduction of the incident wave injection scheme in the Huygens surface A [7]. In any case, as more computational resources become available, larger domains and a tapered wave incident scheme can be introduced with no basic modifications in this method. The absorbing boundary condition (ABC) used in this problem is a perfectly matched layer (PML) [26] formulated with stretched coordinates [27], [28] to match the lossy interior media and provide reflectionless truncation of the computational domain. The stretched coordinate PML formulation is chosen since it can be matched to any arbitrary interior domain, including anisotropic or layered media, without difficulty. In our models, the random media correlation lengths are on the order of a wavelength (resonant regime), so that the fluctuations are directly mapped into the FDTD domain. The permittivity is characterized as where and is a function of position characterizing the random fluctuation. The fluctuation at each position is a Gaussian random variable with zero mean and with correlation function. The generation of is implemented in the Fourier domain by passing a 3-D array of random numbers (with Gaussian distribution and zero mean) through a digital filter whose response corresponds to, the Fourier transform (FT) of. is then the spectral density function of the dielectric fluctuation. We use the following correlation function: where and are the azimuth and vertical correlation lengths, respectively, and is the variance. The correlation function has a Gaussian profile in both the transverse and vertical directions. This correlation function has been used in numerous studies to characterize a number of geophysical media [6], [5], and [29]. In absence of experimental data to support a specific choice of correlation function for particular soil models, a Gaussian function is a preferred choice for its generality and mathematical properties. We note that any general correlation function can be easily incorporated into or FDTD model as well. To create a random medium realization, we begin by defining the 3 D Fourier Transform pair (1) (2) (3) (4) where. To ensure that the dielectric fluctuation will be real, we must enforce the following relation: or equivalently with Let We also assume where and are independent random arrays of Gaussian distribution and zero mean satisfying to preserve the properties of then defined as (5) (6) (7) (8) (9). The dielectric fluctuation is (10) Due to the finite size of the FDTD domain, the generation of random media in this manner has some inherent statistical error (for instance the spatial average permittivity of finite volume realization is not equal to theoretical average, ). This statistical error decreases as the number of correlation lengths in the FDTD volume increase and/or as the variance of the fluctuations decreases. For the media types examined in this paper, the errors are very small (as shown in the next section). III. NUMERICAL RESULTS To first validate the numerical model, we compare the FDTD results against the Born approximation [30]. Given a plane wave incident on the half-space problem defined in Section II, the fields in regions 0 (upper, free-space) and 1 (lower, random medium) satisfy the vector wave equations, respectively (11) (12) where. We can rewrite the vector wave equations in terms of dyadic layered Green s functions, and,as (13) (14) Using the Born approximation to obtain the first-order solution, we replace in the previous integrals by. This approximation is applicable when (low contrast).

4 MOSS et al.: FINITE-DIFFERENCE TIME-DOMAIN SIMULATION 181 Fig. 3. Example random medium permittivity: ^x 0 ^y cross section. Fig. 4. Example random medium permittivity: ^y 0 ^z cross section. The perturbed part of the scattered field in region 0 is then (15) A comparison between the Born approximation results and the numerical results from a 3-D FDTD simulation for the bistatic hh RCS (normal incidence) is shown in Fig. 5. With the application in mind, three types of random media are examined with correlation lengths of and ( being the FDTD cell size), and relative variances of 0.01 and The average relative permittivity is From this figure, we see that the agreement is excellent, particularly for random media with smaller correlation lengths and variances, as expected. The largest error of approximately 3 db occurs for the medium with a correlation length of and a variance of. This is due to the inaccuracy of the Born approximation and the slowly convergent nature of the solution for a relatively larger correlation length. In the following, we study the effects of the random media on the RCS of a buried object. We employ a 3-D FDTD computational domain measuring cells overall, where the random soil comprises cells. This results in a problem with approximately 4.19 million unknowns. For 1000 time steps, it takes approximately 250 CPU seconds to run each realization on an AlphaServer DS20E (667 MHz with 4 GB RAM). The discretization size is chosen to be cm or in free-space, which corresponds to approximately within the soil layer (mean permittivity) for a 500 MHz pulse (center frequency). The source used in this work is a Blackmann Harris pulse [31]. The soil model chosen for this simulation has an average relative permittivity of and a conductivity of S/m, which have both been experimentally determined [5] for a typically dry ground. In this study, the standard deviations of the soil permittivity are chosen to be 10% and 25% of the mean, resulting in variances of 0.01 and , respectively. These values roughly correspond to moisture content (water volume) fluctuations of a few percent, with a mean moisture content approximately less than 5% [9], [22] (note that the soil conduc- Fig. 5. Comparison of the FDTD results and the Born approximation for the three medium types. tivity has a small variance in the models also, but we have determined that its effect on scattering can be neglected in our examples). Fig. 3 depicts the plane cross-section when the correlation length is 20 cells in both directions. Fig. 4 shows the plane cross-section for the same domain with the same correlation length. Three types of random media are considered with different variances and correlations, henceforth referred to as type 1 (, ), type 2 (, ), and type 3 (, ). The conductivity of the media is kept constant. For the types of media in this study, the actual (numerically generated) permittivity means vary slightly (less than a few percent) for each of the different types of media. The actual variances are for type 1, for type 2, and for type 3. The actual material parameters vary slightly due to the statistical error of the finite size random media generation (as mentioned in the previous section). The buried object is placed at a depth such that the random medium scat-

5 182 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 40, NO. 1, JANUARY 2002 Fig. 6. ^h^h bistatic RCS (normal incidence) of random media scattering without object. Fig. 7. object. ^h^h bistatic RCS (normal incidence) of random media scattering with tering amplitude is on the order of the object scattering amplitude. As a result, the RCS from the object will be obscured by the clutter of the random medium. Monte Carlo analysis of the RCS for an ensemble of random media is subsequently performed to extract the bistatic RCS return of the object for normal incidence. The object chosen is a rectangular PEC object, approximately 20 cm 20 cm 10 cm in size, buried m below the surface. 100 realizations are used to obtain the coherent part and the incoherent part of the total RCS, as follows: (16) The coherent component of the RCS extracts (averages out) the scattering of the object alone, which is constant over each realization (recall that the direct layer reflection is purposely not included). The incoherent component is due to the scattering from the random medium fluctuations, which vary over each realization, and produces the variance of the scattered field. Fig. 6 shows a typical (one realization) bistatic RCS (normal incidence) for each type of random medium with no object present. These results show that in this resonant regime, the scattering from the random medium increases as either the variance or the correlation length of the permittivity fluctuations increase. Fig. 7 shows RCS results for the same random medium realization, now with the buried object present. For the type 1 medium, it is apparent that an object is present, because the peak return is almost 10 db greater. For the type 2 medium, the RCS over most scattering angles is slightly greater when the object is present, although for a few scattering angles it is actually less (due to destructive interference). In the type 3 medium, the clutter almost completely obscures the object RCS, and as such the presence of the object has little effect. The bistatic RCS was examined for many random Fig. 8. Ensemble averaging of ^h^h bistatic RCS at 0 incident angle l = l = 31 and =0:01. medium realizations, and these characteristics were observed throughout. In all of the following cases, the bistatic RCS is studied for a TE wave ( scattered fields) at normal incidence. Fig. 8 shows the Monte Carlo result of the object buried in a random medium of type 1. Shown is the coherent average RCS, incoherent average RCS, and the RCS of the object in the homogeneous (mean) medium. The coherent averaging reduces the contribution of the random medium to the RCS, clearly showing the RCS of the object (up to statistical errors). For this random medium, the coherent averaging RCS is very close to that of the object alone, deviating less than 0.5 db in the worst case. The incoherent averaging, on the other hand, shows the variance of the scattered fields of the random medium, in this case a maximum of 32.5 db around 5, or 7.55 db below the scattering from the object itself at that angle. Fig. 9 shows the Monte Carlo result for the object buried in a random medium of type 2. The coherent averaging of the RCS

6 MOSS et al.: FINITE-DIFFERENCE TIME-DOMAIN SIMULATION 183 Fig. 9. Ensemble averaging of ^h^h bistatic RCS at 0 incident angle l = l = 31 and = 0:0625. Fig. 11. Ensemble averaging of ^h^h bistatic RCS at 0 incident angle, l = l =71and =0:0625. Fig. 10. Ensemble averaging of ^h^h bistatic RCS at 0 incident angle, l = l =31and =0:0625, no object present. reconstructs the profile of the object quite well, with a maximum deviation of 1.5 db. The incoherent return is on average about the same as the RCS of the object. In other words, the variance of the RCS contribution from the random medium is approximately that of the RCS from the object alone. With these large fluctuations, determining the presence of the object in this medium is more difficult. The averaging results are not as good as the previous case due to the larger scattered fields from the random medium itself and statistical error. For comparison, Fig. 10 shows the Monte Carlo results for the type 2 medium with no object present. From this figure, we can see that the coherent averaging for the co-polarized field results in an RCS which is 13 to 25 db lower than the incoherent part. The incoherent average of the random medium with the object is between 0.1 db and 0.5 db greater than without the object, due to the interactions between the object and the random medium. With no object present, ideally one could expect the coherent average to be zero, but again the inherent limitations (finite domain size, numerical dispersion, numerical round off error, etc.) of the numerical simulation affect the results. In particular, numerical dispersion affects the randomness in such a way that the statistics of each individual medium are changed in a unique fashion. Note that for each realization, the areas of the lattice that have a higher permittivity are more affected by dispersion than the areas of lower permittivity and hence the overall effect cannot be averaged out: one cannot obtain the scattered field error from the numerical dispersion in the homogeneous half-space alone. Fig. 11 shows the Monte Carlo results for the bistatic RCS for a medium of type 3. In this case, we can see that the incoherent average of the field is approximately 5 db greater than the RCS of the object. As such, detecting the object in this environment could be very difficult. The coherent average of the scattered field does not converge well to the object RCS, with an approximate 2 to 4 db difference in magnitude. The shape is closer to the object RCS at larger observation angles, but diverges greatly between 40 and 20 degrees. However, compared to the RCS of a single random medium realization, or the coherent average without the object present, it is still noticeable that an object is present. Fig. 12 shows the Monte Carlo RCS ensemble averaging convergence for each type of medium, in the backscattering direction. The slower convergence of the type 3 medium is due to the stronger scattering effects of the random medium, and the resulting difficulty in averaging out the coherent part of the RCS. Monte Carlo averaging techniques do well to statistically characterize the random medium scattering effects and can be useful in the design of filters for clutter suppression. However, typical Monte Carlo results (ensemble averages) are theoretical in nature since only one realization can be observed in practice. Therefore, it is important to investigate whether other averaging techniques that can in principle be used in the field to aid in the detection of targets in continuous random media. The closest type of averaging technique one could make would be to shift

7 184 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 40, NO. 1, JANUARY 2002 Fig. 12. Ensemble average convergence, ^h^h normal incidence. Fig. 14. Frequency averaging of TE bistatic RCS at 0 incident angle, l = l =71and =0:0625. Fig. 13. Frequency averaging of TE bistatic RCS at 0 incident angle, l = l =31and =0:0625. the observation beam around the target to obtain distinct (albeit still correlated) realizations [32]. In that case however, the object would be shifting its relative position as the observation beam is moved and this would need to be taken into account in any detection algorithm. Frequency averaging is a technique that can be used in real-world detection problems to discriminate buried objects. The targets must be relatively shallow, due to frequency dependence on penetration depth. The FDTD technique is suited to investigating frequency averaging techniques, but care must be exercised to avoid difficulties at higher frequencies due to numerical dispersion effects. To investigate the potential of frequency averaging in our examples, we average the bistatic RCS over frequency for a single realization of the type 2 and 3 random media considered earlier. The center frequency is 420 MHz, and we extract results from 300 MHz to 540 MHz at 10 MHz intervals (note that variations in medium permittivity with frequency are not considered here). Fig. 13 shows the results for the type 2 random medium, comparing the frequency averaged RCS to the center frequency RCS (in the random medium and the homogeneous medium). The averaged RCS is much closer to the RCS of the object alone, and has effectively removed much of the random medium clutter. The RMS difference between the frequency averaged RCS and the object RCS is 2.47 db, whereas for the center frequency RCS has a RMS difference of 7.7 db. Fig. 14 shows the frequency averaged RCS for the type 3 medium. The averaging in this case does not perform as well due to the larger permittivity fluctuations, but still reduces some of the clutter. The RMS difference from the object RCS is 4.37 db for the frequency averaged RCS, and 8.17 db for the center frequency RCS. Both of these cases have shown that frequency averaging does help discriminate the buried object in our examples. Frequency averaging is difficult to apply to these cases (in the resonant regime), as the higher frequencies have a much stronger scattering response from the random medium, while the lower frequencies show a weaker response from the target. Other techniques such as the angular correlation function (ACF) [19] can also be applied to this kind of problem to aid in target detection under strong clutter conditions. Further studies can be undertaken to determine if this or other techniques are successful in removing the clutter under strong volume-scattering effects. IV. CONCLUSION A 3-D FDTD scheme has been introduced to model electromagnetic scattering from objects below random media. The random medium under consideration was chosen to be an inhomogeneous soil with a spatially fluctuating random permittivity and prescribed correlation function. Limitations on the model imposed by the computational domain size (available computational resources) were discussed. The effect of the volumetric scattering from the random media on the RCS of a buried object was studied. Monte Carlo averaging was performed on three types of random media with and without an object present for bistatic RCS (normal incidence). The coherent averaging was

8 MOSS et al.: FINITE-DIFFERENCE TIME-DOMAIN SIMULATION 185 used to closely reconstruct the RCS profile of the object in a homogeneous medium, and accuracy was better for random media with smaller permittivity variances and correlation lengths (near the resonance regime). The incoherent averaging was used to determine the basic statistical properties of the random medium scattering, e.g., the variance of the random medium scattered fields. Frequency averaging was also performed, and was successful in averaging out part of the random medium clutter in our examples. Further studies of other postprocessing techniques such as ACF may be performed to determine their effectiveness in reconstructing the RCS response under these conditions. REFERENCES [1] A. K. Fung, Scattering from a vegetation layer, IEEE Trans. Geosci. Electron., vol. GE-17, pp. 1 6, Jan [2] K. O Neill, R. F. Lussky, Jr., and K. D. Paulsen, Scattering from a metallic object embedded near the randomly rough surface of a lossy dielectric, IEEE Trans. Geosci. Remote Sensing, vol. 34, pp , Mar [3] F. T. Ulaby, R. K. Moore, and A. K. Fung, Microwave Remote Sensing: Active and Passive. Reading, MA: Artech House, [4] M. Z. M. Jenu and D. H. O. Bebbington, Two-dimensional modeling of optical propagation in turbulent atmosphere, Electron. Lett., vol. 29, pp , Apr [5] R. G. Atkins, Microwave scattering and synthetic aperture radar imaging of targets buried in random media, Ph.D. dissertation, Mass. Inst. Technol., Cambridge, [6] M. Borgeaud, R. T. Shin, and J. A. Kong, Theoretical models for polarimetric radar clutter, J. Electromagn. Waves Applicat., vol. 1, no. 1, pp , [7] A. Taflove, Computational Electrodynamics: The Finite-Difference Time-Domain Method. Norwood, MA: Artech House, [8] K. S. Yee, Numerical solution of initial boundary value problems involving Maxwell s equations in isotropic media, IEEE Trans. Antennas Propagat., vol. 14, pp , Mar [9] J. Bourgeois, A complete three-dimensional electromagnetic simulation of ground-penetrating radars using the finite-difference time-domain method, Ph.D. dissertation, Georgia Inst. Technol., Atlanta, [10] T. J. Cui and W. C. Chew, Fast evaluation of Sommerfeld integrals for EM scattering and radiation by three-dimensional buried objects, IEEE Trans. Geosci. Remote Sensing, vol. 37, pp , Mar [11] H. Brunzell, Clutter reduction and object detection in surface penetrating radar, in Radar 97 Conf., Goteborg, Sweden, [12] K. Demarest, R. Plumb, and Z. Huang, FDTD modeling of scatterers in stratified media, IEEE Trans. Antennas Propagat., vol. 43, pp , Oct [13] D. A. Hill, Electromagnetic scattering by buried objects of low contrast, IEEE Trans. Geosci. Remote Sensing, vol. 26, pp , Mar [14] S. Vitebskiy, L. Carin, M. A. Ressler, and F. H. Le, Ultra-wideband, short-pulse ground-penetrating radar: Simulation and measurement, IEEE Trans. Geosci. Remote Sensing, vol. 35, pp , May [15] T. Dogaru and L. Carin, Time-domain sensing of targets buried under a rough air ground interface, IEEE Trans. Antennas Propagat., vol. 46, pp , Mar [16] K. Demarest, Z. Huang, and R. Plumb, An FDTD near- to far-zone transformation for scatterers buried in stratified grounds, IEEE Trans. Antennas Propagat., vol. 44, pp , Aug [17] F. L. Teixeira, W. C. Chew, M. Straka, M. L. Oristaglio, and T. Wang, Finite-difference time-domain simulation of ground penetrating radar on dispersive, inhomogeneous, and conductive soils, IEEE Trans. Geosci. Remote Sensing, vol. 36, pp , Nov [18] L. Tsang and J. A. Kong, Multiple scattering of electromagnetic waves by random distributions of discrete scatteres with coherent potential and quantum mechanical formulism, J. Appl. Phys., vol. 51, pp , [19] G. Zhang, L. Tsang, and K. Pak, Angular correlation function and scattering coefficient of electromagnetic waves scattered by a buried object under a two-dimensional rough surface, J. Opt. Soc. Amer. A, vol. 15, pp , Dec [20] Y. Miyazaki, J. Sonoda, and Y. Jonori, Statistical analysis of electromagnetic scattering of buried objects in random media using FDTD method, Trans. Inst. Elect. Eng., C, vol. 117-C, pp , Jan [21] S. G. Tanyer, M. Karaman, and I. Ozturk, Analysis of wave propagation in inhomogeneous media using FDTD method and its applications, in Int. Conf. Mathematical Methods in Electromagnetic Theory, Kharkov, Ukraine, 1998, pp [22] W. R. Scott and G. S. Smith, Measured electrical constitutive parameters of soil as functions of frequency and moisture content, IEEE Trans. Geosci. Remote Sensing, vol. 30, pp , May [23] C. H. Chan, S. H. Lou, L. Tsang, and J. A. Kong, Electromagnetic scattering of waves by random rough surface: A finite-difference time-domain approach, Microw. Opt. Technol. Lett., vol. 4, no. 9, pp , [24] F. D. Hastings, J. B. Schneider, and S. L. Broschat, A Monte-Carlo FDTD technique for rough surface scattering, IEEE Trans. Antennas Propagat., vol. 43, pp , Nov [25] C. D. Moss, Finite-difference time-domain simulation of electromagnetic scattering from objects under random media, M.S. thesis, Mass. Inst. Technol., Cambridge, [26] J. P. Berenger, A perfectly matched layer for the absorption of electromagnetic waves, J. Comput. Phys., vol. 114, pp , [27] W. C. Chew and W. H. Weedon, A 3-D perfectly matched medium from modified Maxwell s equations with stretched coordinates, Microw. Opt. Technol. Lett., vol. 7, pp , Sept [28] F. L. Teixeira and W. C. Chew, A general approach to extend Berenger s absorbing boundary condition to anisotropic and dispersive media, IEEE Trans. Antennas Propagat., vol. 46, pp , Sept [29] A. K. Fung and F. T. Ulaby, A scatter model for leafy vegetation, IEEE Trans. Geosci. Electron., vol. 16, no. GE-4, pp , Oct [30] L. Tsang, J. A. Kong, and K. H. Ding, Scattering of Electromagnetic Waves: Theories and Applications. New York: Wiley, [31] F. J. Harris, On the use of windows for harmonic analysis with the discrete Fourier transform, Proc. IEEE, vol. 66, pp , Jan [32] K. O Neill, Exploration of innovative radar sensing schemes for subsurface object detection, in Proc. IGARSS 97, vol. 3, 1997, pp C. D. Moss (SM 01) received the B.S. degree from the University of Alberta, Edmonton, AB, Canada, in 1996, and the M.S. degree in electrical engineering from the Massachusetts Institute of Technology (MIT), Cambridge, in He is currently pursuing the Ph.D. degree in electrical engineering at MIT. He was an Engineer at Raytheon in the Microwave Systems Department from 1997 to 1998, where he was mainly involved in MMIC design. Since 1998, he has been a Research Assistant at MIT in the Center for Electromagnetic Theory and Applications, and in Lincoln Laboratory. His research interests include both analytical and computational methods in electromagnetics. F. L. Teixeira (M 00) received the B.S. and M.S. degrees in electrical engineering from the Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Brazil, in 1991 and 1995, respectively, and the Ph.D. degree in electrical engineering from the University of Illinois at Urbana-Champaign, Urbana, IL, in From 1999 to 2000, he was a Postdoctoral Research Associate with the Research Laboratory of Electronics, Massachusetts Institute of Technology (MIT), Cambridge, MA. Since 2000, he has been an Assistant Professor at the ElectroScience Laboratory (ESL) and Department of Electrical Engineering, The Ohio State University, Columbus, OH. His current research interests include analytical and numerical techniques for wave propagation and scattering modeling in communication, sensing, materials, and devices applications. Dr. Teixeira was awarded the Raj Mittra Outstanding Research Award from the University of Illinois, a 1998 IEEE MTT-S Graduate Fellowship Award, and paper awards at the 1999 USNC/URSI National Radio Science Meeting (Boulder, CO) and at the 1999 IEEE AP-S International Symposium (Orlando, FL). He was the Technical Program Coordinator of the Progress in Electromagnetics Research Symposium (PIERS) in Cambridge, MA, in 2000.

9 186 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 40, NO. 1, JANUARY 2002 Y. Eric Yang (M 93) was born in Taichung, Taiwan, R.O.C. He received his Ph.D. degree in electrical engineering from the Massachusetts Institute of Technology (MIT) in He was an electronics instructor of the Taiwanese Navy from 1981 to After his Ph.D. studies, he became a Postdoctoral Associate (1989) and then Research Scientist ( ) in the Research Laboratory of Electronics at MIT, where he conducted research in electromagnetic scattering and remote sensing, with emphasis on computer simulation applications. He led a multi-year effort to develop EMSARS, a graphical user interface driven radar image simulation software, for U.S. Department of Defense clients. He also developed simulation models to analyze radionavigation system performance and interference immunity for U.S. and U.K. government agencies. He made great contributions to the Microwave Landing System demonstration programs initiated by the Department of Transportation during the period Since December 1999, he has been with the Rannoch Corporation, focusing on avionics and aviation systems research and development. Dr. Yang s research interest includes the areas of wave propagation in microelectronic circuits, computational techniques, electromagnetic interference in electronic systems, wave propagation, scattering, and radar imaging simulation. He is a Member of The Electromagnetics Academy, the IEEE, and Sigma Xi. He received the triennial Henry George Booker Award for Young Scientist by the URSI US National Committee (1990). Jin A. Kong (S 65 M 69 SM 74 F 85) is the President of The Electromagnetics Academy and Professor of electrical engineering at the Massachusetts Institute of Technology, Cambridge, where he is also the Chairman of Area IV on Energy and Electromagnetic Systems in the Department of Electrical Engineering and Computer Science and Director of the Center for Electromagnetic Theory and Applications in the Research Laboratory of Electronics. His research interest is in the area of electromagnetic wave theory and applications. He has published over 30 books including Electromagnetic Wave Theory (Wiley 1975, 1986, 1990; EMW since 1998) and over 600 refereed journal articles, book chapters, and conference papers. He has been Principal Investigator for more than 100 grants and contracts from various government agencies and industry. He has served as Consultant, External Examiner, and Advisor to industry, academia, national governments, and the United Nations. He has been Reviewer for many journals, book companies, and government agencies. He is Editor for the Series in Remote Sensing (New York: Wiley), Editor-in-Chief of the Journal of Electromagnetic Waves and Applications (JEMWA) and Chief Editor for the book series Progress In Electromagnetics Research (PIER). Dr. Kong has served as Session Chairman, Organizer, and member of advisory and technical program committees for numerous international and national conferences and symposia, including serving as Chairman for the Progress In Electromagnetics Research Symposium (PIERS) since In 2000, he received the S. T. Li Prize and the Distinguished Achievement Award from the IEEE Geoscience and Remote Sensing Society.