Validation of a 3-D Near-Field ISAR Imaging Technique with Far-Field RCS Extraction by Means of a Hybrid GO-PO/PTD Ray Tracing

Algorithm
Thomas Vaupel, Frank Weinmann
Research Institute for High Frequency Physics and Radar Techniques, FGAN e.V. Neuenahrer Str. 20, 53343 Wachtberg, Germany
vaupel@fgan.de, weinmann@fgan.de Abstract This paper deals with a 3-D inverse synthetic aperture radar (ISAR) imaging technique using a cylindrical scanning plane in the near-field of the radar target, where the resulting 3-D images are displayed on sheets perpendicular to the scanning cylinder axis. Besides of a very detailed scattering center analysis of targets with arbitrary shapes, the method allows in a second step the extraction of the far-field radar cross section (RCS) of the targets without the need of long range measurement set-ups or costly compact-range arrangements.

of the imaging and extraction methods as well as for the raytracing algorithms. II. FORMULATION The setup for the acquisition of the near-field scattering information by a cylindrical scanning technique is given in Fig. 1.
y

I. INTRODUCTION In the last years, two-dimensional near-field ISAR imaging techiques have been intensively used for the characterization of radar targets [1, 2, 3]. A very important feature of these techniques is given by the ability to extract the far-field RCS from the image information. Thus, the need for costly longrange or compact-range measurement set-ups [4] can be circumvented by these methods. However, for a reasonable scattering center representation and accurate far-field extraction, the far-field condition must be fulfilled with regard to the height of the objects restricting the application of these two-dimensional algorithms to more or less flat objects such as aircrafts or elongated air targets and aggregates. Therefore near-field ISAR techniques with a threedimensional processing are necessary for a detailed and accurate characterization of objects with arbitrary shape. Thus, in addition to the angular discretization, a vertical scanning must be provided by moving the transmitting and receiving antennas to capture the vertical extent of the target. Three-dimensional near-field imaging techniques can be already found in [5, 6], but not with the aim of a further extraction of far-field information. To get this information, in a first step a reliable control of the measurement and imaging process is provided by the derivation of a sampling theorem for the height discretization. Afterwards the image information is computed with two-dimensional convolution techniques as well as a polar reformatting method. The farfield information is then extracted by a summation over fast cyclical convolutions. For a test of the imaging and extracting techniques, the scattered fields can be advantageously calculated by using appropriate simulation tools, e.g. high-frequency methods based on ray-tracing. This procedure allows for the validation

top view
1

R
imax

R, E

z
R, E
2

d ( , , z h)
y
z h

Fig. 1 Setup for the cylindrical near-field scanning techique.

The target is positioned on a turntable which is rotated by an angle , whereas the transmitting and receiving antennas are positioned in different heights h. The quasi-monostatic arrangement can also be generalized to a bistatic configuration. If we assume that the target has a reflectivity distribution ( , , z ) consisting of independent scattering centers, then the scattered electric field can be given by the following integral representation

691

Es ( f , , h) =
i max 2 z = hobj / 2

Csys

( , , z)

= 0 0 z = hobj / 2

e d 2 ( , , z h)

f j 4 d ( , , z h ) c

(1)

content of the electric field with regard to the vertical sampling. From this, the maximum sampling intervall can be derived: h <
2 c R 2 + hobj

GR ( f ,1 ) GE ( f ,1 ) dz d d

4 f max hobj

(9)

(2) R 2 + 2 2 R cos( ) + ( z h) 2 in case of a quasi-monostatic arrangement. Furthermore the frequency dependent gain functions GR and GE of the receiving and emitting antenna can be optionally considered. To get the reflectivity contribution from the scattered nearfield, an inverse integral transformation is performed ( , , z) = f 2 h / 2 (3) Es ( f , , h) ( f , , , z h) dh d f df
2 obj

with d ( , , z h) =

The focusing operator is an even function with regard to z. If the scattered field is sampled at N E sampling points in zdirection covering the whole object height, the focusing operator must be computed at N = 2 N E 1 sampling points to evaluate the convolution at N E points. For a fast convolution without overlapping errors, the FFT-length must be larger than 3N E 1 points. However, since the computation of N E image planes is sufficient to cover the whole object, some located overlapping errors in the convolution result can be tolerated. It can be shown that the FFT-length must be only 2 N E 1 to evaluate N E image planes without overlapping errors. After evaluating eq.(5), the image information is still given in a polar format. To get the familiar Cartesian representation, selected image planes are treated with a polar reformatting algorithm. As outlined in [1], a reformatting of the complete complex image information is not possible with high accuracy, whereas the absolute value can be reformatted without any visual errors, i.e., ( , , zi ) polar reformatting ( x, y, zi )

f = f1 0 h = hobj / 2

using the focusing operator ( f , , , z h) =

f j 4 d ( , , z h ) (4) d 2 ( , , z h) e c GR ( f ,1 , 2 ) GE ( f ,1 , 2 ) For the efficient evaluation of this inverse transformation, the two-dimensional convolution relationship with regard to and z between the electric field and the focusing operator is utilized, leading to ( , , z) = f 5) IFFT , z FFT , z [ ( f , , , z )] FFT , z [ Es ( f , , z )] f df
2

f = f1

For a reliable evaluation of eq.(4) and (5) without aliasing effects, appropriate guidelines for the discretization must be fulfilled. For this goal, the electric field of N ideal point scatterers at locations ( p , p , z p ) is considered. Using eq.(1)
we get: Es ( f , , z ) p
p =1 N
p p p e 2 d ( p , p , z z p )

is performed for the selected zi . For the extraction of the scattered far-field, we can write for eq.(2): d ( , , z h) =
2 ( z h) 2 2 (10) R 1 cos( ) + 2 + R R R2 2nd order terms R cos( ) Since we can write ( , , z ) = ( , , zi ) ( z zi ) , we get with eq.(10) in the exponential of eq.(1) and the filtering property of the delta-function:
2

j 4 ( f / c ) d ( , , z z )

(6)

For the discretization with regard to we get with a similar derivation as in [1, 2] for the maximum angular interval of the electric field sampling:

<

2 c R 2 + p ,max

E ( f , ) =
FF s

4 f R c

4 f max R p ,max

(7)
zi

R2
i

j 4 f cos( ) c

with p ,max the maximum radius of the object and f max the maximum operation frequency. With the first derivative of d ( , , z ) in the exponentials of eq.(6), we get a rule for the height discretization. This leads to z zp hobj d ( , , z ) (8) |max = |max = 2 2 2 z R + (z zp ) R + hobj

i max
=0

(11)
d

( , , z ) e
0
j 4 f R c

one-dimensional cyclical convolution in

Thus, the far-field is efficiently computed by

E ( f , ) =
FF s

R2

(12)

with hobj the overall height of the object. The maximum of this derivation is proportional to the maximum spectral

f j 4 cos( ) ] d IFFT FFT [ ( , , zi )] FFT [e c zi = 0

i max

directly from the image information in polar format.

692

III. APPLICATIONS A. Ideal Point Scatterers The image processing and far-field RCS extraction can be traced and controlled very precisely with distributions of ideal point scatterers, where the far-field RCS or the scattering coefficient ( f , ) can be computed exactly using eq.(6) and a far-field approximation of the exponential. Six point scatterers are used with the following coordinates and scattering coefficients:
TABLE I DISTRIBUTION OF POINT SCATTERERS

The exact far-field RCS and the extracted far-field RCS are nearly congruent whereas the near-field RCS exhibits significant deviations from the far-field results.

p
1 2 3 4 5 6

x p (cm) -3 0 2 0 -3 0

y p (cm) 0 -4 0 3 0 0

z p (cm) -30 -15 0 0 0 15

p (m)
2.0 1.0 1.0 2.5 1.5 1.5
0

-6 -5 -4 -10 -15 -20 -25 -30 -35 6 -6 -4 -2 0 2 4 6 -40

-2

y (cm)

dB

z/cm

x/cm

y/cm

Fig. 3 Three-dimensional processing of the scatterer distribution of Table 1.

x (cm)
Fig. 2 Two-dimensional processing of the scatterer distribution of Table 1.
20

The frequency range is chosen to f = 18-25 GHz, the distance 15 R to 5 m. According to eq.(9), this leads to a vertical discretization of h = 2.5cm and NE = 25 sampling points to cover the vertical extension of the scatterer distribution. Fig. 2 10 shows the two-dimensional image generated with one of the methods in [1, 2]. In this case, only 5 scatterers can be 5 distinguished, since scatterer 1 and 5 have the same lateral coordinates. Fig. 3 shows the image result of the three0 dimensional processing. Only five image planes at z = -30, -15, 0, 15 and 30 cm are treated with the reformatting algorithm -5 and are subsequently displayed. Now all scatterers can be distinguished together with their different scattering coefficients. Fig. 4 shows the results for the near-field RCS -10 0 50 100 150 200 250 300 350 (blue), the exact far-field RCS (green) and the extracted far / degree field RCS derived with eq.(12) (red). For the calibration and restauration of the results of the extracted RCS, a single point Fig. 4 Comparison of near- and far-field RCS for the scatterer distribution of scatterer with = 1 m in the center of the coordinate system Table 1 and Fig. 3. was processed in the same way as the distribution of Table 1.

693

RCS / dBsm

B. Processing of data generated with the FARAD Ray Tracing Tool FARAD (FAst Ray-tracing Algorithm with ray Density normalisation) is a ray tracing software designed for the electromagnetic modelling of arbitrarily large objects. The tool has been developed at FGAN-FHR and is based on discrete rays, which are used for finding the relevant propagation paths. These rays are traced according to the wellknown Shooting and Bouncing Rays (SBR) technique, which allows for a practically unlimited number of reflections. The SBR approach is combined with the Physical Optics (PO) and Physical Theory of Diffraction (PTD), which are used to calculate scattered field contributions at each intersection of a ray with the object [7]. An additional implementation of the Uniform Theory of Diffraction (UTD) has proved to become an important enhancement when multiple interactions with the object are relevant. This is of special interest for the accurate modelling of cavities, where rays might be diffracted at the aperture and propagate into the cavity. According to the UTD implementation, these diffracted rays can be traced further through the scenario and generate additional PO and PTD contributions, which might be relevant for the total scattered fields. Thus, a hybrid GO/UTD-PO/PTD simulation code has been developed, with the laws of GO and UTD (i.e. law of reflection and diffracted rays on the Keller cone) serving for calculating the geometrical propagation paths while PO and PTD are used for the calculation of field contributions. Further enhancements of the code include the modelling of arbitrary surfaces by importing reflection and transmission coefficients from tables and the adequate treatment of curved surfaces. Due to the fast generation of results for different frequencies and aspect angles, scattering studies which generally require enormous simulation efforts are possible within a reasonable amount of time. Thus, for example, the ray tracing tool is able to provide data for the generation of one-dimensional range profiles of the object as well as twodimensional images of the object, which are essential for the analysis of scattering centers and optimization concerning Low Observability (LO). In the scope of this paper, only the GO-PO/PTD part of the ray tracing tool is applied to a generic scattering object, which is made up of a set of cuboids (Fig. 5). A special feature of the ray tracing tool is the ability to perform near-field calculations by simply placing the ray source and observations points in the near-field of the object. For this generic scattering object, monostatic scattering simulations are performed at the distance R = 9 m. As the height of the object is 65 cm, only 7 sampling planes are required to capture the vertical characteristics of the object. Each of these sampling planes requires 720 sampling points for the angular discretization. Fig. 6 depicts the two-dimensional processing with the vertical center cuboid and the additional horizontally aligned cuboids. Note that the visualization is mirrored at the x-axis compared to the geometry of the object in Fig. 5. The threedimensional processing of the object (Fig. 7) shows the vertical setup of the object with remarkable accuracy, even

Fig. 5 Generic scattering object..

0 -5 -10 -15 -20 -25 -30 -35 -40

-50 -40 -30 -20 -10

cm

0 10 20 30 40 50 -50

dB

cm

50

Fig. 6 Two-dimensional processing of the scatterer distribution for the object displayed in Fig. 5.

for the low center frequency of 8 GHz. As expected due to the coarse resolution, each of the intermediate planes at z = 10 cm and z = 10 cm shows two of the additional cuboids sticking out of the vertical center cuboid. Subsequently, the far-field RCS was extracted from the threedimensional image information. The results for the extracted far-field RCS are given in Fig. 8 (red curve) together with the far-field results of the GO-PO/PTD algorithm (green curve) which are derived by choosing an antenna to object distance of R = 1000 m. Furthermore the near-field RCS was computed by means of the scattered near-field in the plane z = 0 (blue curve). With regard to the extracted far-field RCS (red) and the far-field results of the GO-PO/PTD algorithm (green), a good agreement for the mainlobes and other areas of the azimuth angle can be observed. But e.g. the sidelobes directly around the mainlobes are not in good agreement and peaks can be found in the results of the GO-PO/PTD algorithm which correspond with dips of the extracted results and vice versa. These deviations are now under further investigation.

694

cm

IV. CONCLUSIONS For the detailed characterization of arbitrary radar targets, a 3 -D near-field ISAR imaging technique with cylindrical scanning is presented together with precise guidelines for the height and angle discretization. The image information is derived in a polar format with high efficiency due to a twodimensional fast convolution technique and is subsequently displayed on sheets perpendicular to the vertical axis in Cartesian format using a reformatting algorithm, leading to a very detailed target representation. The extraction of the farfield RCS is performed in a second step based on the polar image information and restoration techniques. With this feature, the far-field RCS of arbitrary targets can be determined without costly long range or compact range measurement arrangements. The techniques are validated with the help of point scatterer distributions and a hybrid GO/UTDPO/PTD simulation code. REFERENCES
[1] Vaupel, T., Eibert, T. F., Comparison and Application of Near-Field ISAR-Imaging Techniques For Far-Field Radar Cross Section Determination, IEEE Trans. Antennas and Propagation, Vol. 54, No. 1, Jan. 2006, pp. 144-151. Vaupel, T., Eibert, T. F., Far-Field Radar Cross Section Determination using Near-Field ISAR-Imaging Techniques, 6th European Conference on Synthetic Aperture Radar (EUSAR),Dresden, 16.--18. May 2006. A. Broquetas, L. Jofre, A. Cardama, Spherical wave near-field imaging and radar cross-section measurement, IEEE Trans. Antennas and Propagation, Vol. 46, No. 5, pp. 730-735, May 1998. A. Broquetas et al., A compact system for radar cross section measurement and imaging up to 40 GHz, in Proc. Jina, Nice, France, 1990, pp. 596-599. J. Fortuny, An efficient three-dimensional near field ISAR algorithm using the method of stationary phase, IEEE Trans. Aerospace Elect. Syst., vol. 34, pp. 1261-1270, Oct. 1998. J. M. Lopez-Sanchez, J. Fortuny, 3-D Radar Imaging using range migration techniques, IEEE Trans. Antennas and Propagation, vol. 48, pp. 670-684, March 2000. F. Weinmann, Ray Tracing with PO/PTD for RCS Modeling of Large Complex Objects, IEEE Trans. Antennas Propagat., vol. 54, pp. 1797-1806, June 2006.

cm

cm

[2]

Fig. 7 Three-dimensional processing of the scatterer distribution for the object displayed in Fig. 5. [3]
30

[4]
20

[5] [6] [7]

RCS / dBsm

10

-10

-20

-30

50

100

150

200

250

300

350

/ degree
Fig. 8 Comparison of near- and different far-field RCS results for the generic scattering object of Fig. 5. Blue: Calculated near-field RCS. Red: Extracted far-field RCS. Green: Calculated far-field RCS.

695