Proceedings of the Institute of Acoustics

 

Reverse-path multistatic SAR for moving target detection in clutter

 

D. Andre, Centre for Electronic Warfare, Information and Cyber, Cranfield University, Defence Academy of the United Kingdom, SN6 8LA, UK
F. M. Watson, Department of Mathematics, The University of Manchester, Oxford Road, Manchester, M13 9PL, UK

 

1 INTRODUCTION

 

Synthetic Aperture Radar (SAR) is a side looking remote sensing radar imaging mode. Due to the SAR image formation process being based on the assumption of a stationary scene, it is well known and observed that target motion causes defocusing and displacements¹. In particular, targets with motion towards the radar, i.e. moving in a direction perpendicular to the radar trajectory, and with a velocity which has a constant radar range direction component only, appear both focused and displaced in azimuth in corresponding SAR images. While blurring due to azimuthal motion may be focused and used to estimate the azimuthal component of velocity², the displacement can present a difficulty to image analysts, as it is then unclear where the target actually is or what its radial speed is, especially when its signature is below the ground clutter signal level. Note that the direction of this displacement depends on both whether the target is incoming or outgoing from the radar, and which direction the radar itself is moving in, which we later exploit in this paper.

 

Multistatic SAR is often considered as a means to improve image resolution and target recognition³, due to the multiple bistatic source/receiver configurations increasing the extent of k-space covered. More generally, multistatic configurations may provide enhanced spatial, temporal (range), or spectral (Doppler) resolution⁴. As a special case, we propose a multistatic configuration which provides no benefit to the spatial image resolution, but does provide motion resolution of moving targets. This is achieved by the use of a reverse-path multistatic radar configuration, consisting of one co-located transmitter/receiver travelling with azimuth angular velocity ω_M about the scene centre, and a second receiver travelling in the opposite direction with angular velocity ω_R = -3ω_M about the scene centre, at same grazing angle and with the centre of each aperture approximately co-located. The monostatic and bistatic apertures then cover close to the same extent in K-space⁵, but have obtained the collection in a different order from one-another during the simultaneous collections. This results in a pair of images which have SAR moving target displacements in opposite directions, and are also well suited to coherent subtraction of ground clutter, allowing detection and imaging of range-moving targets against strong background clutter. Range direction target speed can then be readily estimated either by inspection or through windowed autocorrelation of the difference image in the cross-range direction.

 

An equivalent data collection can be carried out by having the transmitter at a greater range than the bistatic receiver, reducing its speed required to create the same equivalent mono-static aperture. An operationally relevant collection might be a stand-off platform with high-powered radar, with a stand in slower-moving receive-only UAV. This can be generalised further to other multi-static configurations in which the extent of k-space coverage is approximately the same, but the azimuth order in which it is collected is different, demonstrating a special case of gaining temporal resolution, instead of finer spatial resolution, using multistatic SAR configurations. This approach differs from the Ground Moving Target Indication (GMTI) mode^6,7, in that GMTI used several radar phase centres and provides only coarse azimuth real radar beam-based resolution detections / indications. Generally, these coarse GMTI detections are overlayed on conventional high resolution SAR images, thus providing useful information of where moving targets are, and their speed, but not of a sufficient resolution to allow visual identification.

 

We present here numerical experiments into the effectiveness of the method in detecting weak targets against strong speckle clutter. The simulation is described in section 2, the results in section 3, and finally the discussion and conclusion in section 4.

 

2 SIMULATION

 

A SAR simulation was developed to demonstrate the multistatic reverse-path technique. The simulation employs a scattering scene comprising a collection of point-scatterer targets, and calculates the complex scattering amplitude from the scene, given the transmitter and receiver positions, and the radar frequencies sampled. The simulation is performed for bistatic synthetic aperture radar trajectories, with some of the scatterers in the scene – the target – moving during the radar collection.

 

The SAR geometry is shown in Figure 1, with the target scene at the origin, a monostatic collection trajectory at 100 km range and 20º grazing angle (red) moving to the right (positive x-direction), and a receiver collection trajectory at 10 km range and 20º grazing angle (cyan) moving to the left (negative x-direction). The sensed frequency range is 300Mhz bandwidth with centre frequency of 5 GHz. The monostatic collection aperture extent is 3º, whereas the receiver aperture extent is 9º, travelling in the opposite direction at three times the azimuth angular speed as the monostatic collection. The two apertures are horizontal and centred about the same direction from the target scene.

 

The two radar collections – monostatic and bistatic – cover approximately the same region of K-space within the same time period, but in opposite azimuthal directions, which is the crucial requirement for the proposed approach. When considering the bistatic collection, it is helpful to consider the bistatic equivalent monostatic phase centre motion, which in this case can be roughly approximated at the midpoint between transmitter and receiver positions. The achieved SAR resolution in ground range and cross-range is 0.53 m and 0.57 m respectively.

 

 

Figure 1: The SAR geometry: the target scene is at the origin; the monostatic radar platform trajectory is seen in red at 100 km range traversing to the right; the receiver radar platform trajectory is seen in cyan at 10 km range traversing to the left.

 

The target scene is presented in Figure 2. It is comprised of point scatterers, with a random distribution of scatterers on the ground plane in a 10 x 10 m square, with a simple model vehicle over it 2m in height. The ground scatterers are distributed with a sufficiently fine density to ensure that several of them will be found within the estimated SAR cell resolution size, which together with their randomized positioning, ensures that they will manifest in the corresponding SAR image as developed speckle clutter.

 

The model vehicle is translated at constant speed towards the radar in the positive y -direction during the SAR collection by a total of 15 cm in ground-range, with the blue dots indicating its original position, and the red its final position. To provide a challenging scenario, the scattering amplitude of the target point scatterers is set to be 1/10^th of the scattering amplitude strength of the ground clutter scatterers.

 

 

Figure 2: The target scene is composed of point scatterers, comprised by a random distribution of scatterers on the ground plane, with a simple model vehicle over it. The model vehicle is translated forwards in the positive y-direction towards the radar during the SAR collection, with the blue dots indicating its original position, and the corresponding red dots its final position.

 

With regards the speeds of the radar platforms and target, the relative motions between these are set in the simulation, hence when one of them is set, then all the others are fixed. Hence if the stand -off monostatic platform speed is assumed to be 900 km/h, then the receiver platform, which in the current simulation is 10 times closer to the target scene (10 km) and is necessarily three times faster in terms of azimuth angle speed, its speed is must then be 900/10*3 = 270 km/h. Similarly, it is noted that if the receiver platform were instead at 1 km range, then its required speed would be 900/100*3 = 27 km/h. With its 900 km/h speed, the monostatic azimuth aperture of 3º will be traversed in 19.7 s, setting the multistatic SAR collection time. This implies the set target displacement of 15 cm occurs with a speed of 7.6 mm/s.

 

3 RESULTS

 

The Backprojection algorithm⁵ was implemented for image formation, and the monostatic and bistatic SAR images are presented Figure 3. These images show the strong ground clutter which manifests itself here as developed speckle. Within these images the target signature is also known to be present, however it is not clear where it is due to the strength of the clutter signature. It is noted that because the monostatic and bistatic collections cover approximately the same region of K-space, it is the case that both collections are coherent, giving rise to the high level of similarity in speckle pattern between both images⁸. The similarity in the coherent speckle will allow the moving target signature to be extracted.

 

Figure 3: Monostatic (a) and bistatic (b) SAR images of the scene with moving target.

 

The coherent sum and subtraction of the monostatic and bistatic images are presented in Figure 4, with absolute intensity scales to allow cross comparison of scattering strength between scattering signatures. The coherent sum dominant speckle pattern clutter looks much the same as that in the SAR images in Figure 3, however the coherent subtraction image reveals two target signatures which were previously overwhelmed by the clutter. The two signatures can be seen to be very similar to each other, and can be understood as either left or right cross-range translated versions of the single moving target, where the translation is due to the target radar range motion during the SAR collection. The left-hand signature is associated with the bistatic SAR collection, and the right-hand signature is associated with the monostatic SAR collection.

 

 

Figure 4: The coherent sum (a) and subtraction (b) of the monostatic and bistatic images previously presented in Figure 2. Non-normalised intensity scales are employed and the x-direction corresponds to cross-range.

 

In practice, the cross-range separation between the two signatures can be used to estimate the target speed during the SAR collection. In the monostatic collection, the SAR image moving target cross - range translation, SM, is given by¹

 

 

where V_T is the target speed, ω_M and V_M are the monostatic radar platform angular velocity z component about the scene centre and speed respectively, and R_Mg is the monostatic sensor ground range to the target. Inputting the simulation values into (1) gives a displacement of 2.86 m.

 

For the receiver component of the bistatic collection, for k-space overlap with the monostatic collection, we have imposed the condition

 

 

where ω_R is the bistatic receiver sensor angular velocity z-component.

 

Hence, for the bistatic collection, the SAR image moving target cross -range translation, S_B, is given by

 

 

where (2) is used, and the factor of 2 has entered due to the consideration of the bisecting / bistatic direction (or alternatively the equivalent monostatic aperture).

 

Hence the distance between the two displaced signature separations is 2SM = 5.73 m, which is in close agreement with the separation observed in Figure 4b).

 

With a view towards automating the process of high-resolution moving target detection in this multistatic SAR mode, the autocorrelation of the complex valued subtraction image gives the result displayed in Figure 5. The central brightest peak corresponds to the central autocorrelation value, and the second brightest peaks, which are located at the far left and right, correspond to the cross - range shifted signature detections. These secondary peaks have cross-range values of ± 5.7 m, which are in good agreement with the expected displacement values calculated with (1) and (3) of 5.73 m.

 

This shows that if such a multistatic SAR measurement were conducted of a target moving towards the radar in the presence of strong clutter, that it may be possible to detect the target and determine its speed from the SAR data, and obtain a fine resolution SAR signature of the target which would aid in its recognition. Having determined the nature of the signature, the position of the target is then known to be halfway between the two displaced signatures.

 

 

Figure 5: Autocorrelation of the complex coherent subtraction SAR image seen in Figure 4b), with second brightest peaks with cross-range values of ± 5.7 m.

 

4 DISCUSSION AND CONCLUSION

 

A multistatic SAR mode has been proposed to detect moving targets in SAR images in the presence of strong clutter, providing the moving target signature with the SAR resolution. It is emphasized that even without the presence of clutter, it may not be possible from a single SAR image of a target alone to determine whether its signature is correctly positioned in the image or not, as its cross -range position is dependent on its radar range speed¹. The proposed technique would allow the determination of whether the target is in motion or not, and would provide its correct location. In comparison, GMTI is a conventional non-imaging mode which can also detect moving targets, providing their radar range speeds, however its resolution is based on the real beam size, and is therefore very low at large stand-off ranges precluding recognition^6,7.

 

To add robustness and versatility to the approach, it is envisaged that information from several receiver platforms could be combined, in a higher order multistatic sensing mode, providing multiple moving target signatures with which to perform detection and imaging. It is also possible that such a mode could be arranged to resolve the situation where the target has a motion component in a direction which is not aligned with the radar range direction. In this situation, the target signature can be out of focus¹, however if the receiver trajectory can be arranged appropriately, it is envisaged that the bistatic SAR signature can be brought back into focus. This technique might also be combined with other methods which focus moving targets², making full use of both the multiple bistatic images themselves, as well as the coherent subtraction images, to detect and finely resolve moving targets.

 

It is noted that the physics model under consideration assumes omnidirectional scattering, so that there is an equivalence in Radar Cross Section (RCS) between aligned monostatic and bistatic geometries. In practice however, RCS may vary substantially between monostatic and bistatic geometries, depending on the scatterers⁹. For example, large metal corner retroreflectors such as trihedrals scatter brightly in monostatic geometries, but not for sufficiently large bistatic angles. Metal spheres on the other hand maintain the same RCS under varying bistatic angle. Hence the region of applicability of this mode will depend on several factors, including the type of scatterer, the radar frequency and the size of the bistatic angle. For the scenario simulated in this paper, the largest bistatic angle attained in the mode is 6º, which for 5 GHz may be considered small for common scenes of interest, so that it is likely that bistatic and monostatic RCS will be similar.

 

It is also noted that through laboratory-based bistatic SAR measurements of rough terrain, it has been found that even with substantial K-space overlap, coherence between bistatic SAR images of scenes with developed speckle is dependent on the difference between bistatic angles, where the greater the difference, the lower the coherence¹⁰. This could have negative implications for the coherent subtraction employed in the proposed mode where the bistatic angle is sufficiently large.

 

Finally, it is noted that a possible practical scenario was demonstrated consistent with a large standoff airborne monostatic SAR platform illuminating a target scene, and with a close-in and slower receiver platform, which could for example take the form of a UAV drone-based sensor, which would likely be more covert and possibly expendable¹¹.

 

Upcoming work will involve demonstrating this multistatic mode with measured data obtained at the Ground-Based SAR laboratory at Cranfield University¹².

 

5 REFERENCES

 

1. W. G. Carrara, R. S. Goodman, R. M. Majewski, “Spotlight Synthetic Aperture Radar, Signal Processing Algorithms”, Arthech House, INC., 1995
2. F. M. Watson, “Focusing dynamic single-channel synthetic aperture radar video with optical flow-informed reconstruction”, IET Electronic Letters, Volume58, Issue,25, December 2022
3. D. Blacknell, D. Muff, M. Nottingham and C. Stevenson, "Radar Automatic Target Recognition Based on Disjoint Multi-Static K-Space Coverage," 2018 International Conference on Radar (RADAR), Brisbane, QLD, Australia, 2018, pp. 1-5, doi: 10.1109/RADAR.2018.8557271.
4. Margaret Cheney and Brett Borden, “Imaging moving targets from scattered waves,” Inverse Problems 24 035005, 2008
5. C. V. Jakowatz, Jr., D. E. Wahl, P. H. Eichel, D. C. Ghiglia, P. A. Thompson, “Spotlight-Mode Synthetic Aperture Radar: A Signal Processing Approach”, Kluwer Academic Publishers, 1996
6. J. R. Guerci, “Space-Time Adaptive Processing for Radar, second edition”, Artech House, 2015
7. G. W. Stimson, “Introduction to Airborne Radar, second edition”, Scitech Publishing Inc., 1998
8. D. Blacknell, D. Andre, A. Bennett, “SAR coherent change detection”. In: New methodologies for understanding radar data, p. 151-196, SciTech Publishing, 2021
9. E. F. Knott, J. F. Shaeffer, M. T. Tuley, “Radar Cross Section, second edition”, Artech House, 1993
10. A. Hagelberg, D. Andre, M. Finnis, “Laboratory Bistatic SAR CCD Investigation”, Electronic Letters, 59, (15) Article No. e12913., 2023
11. N. J. Willis, H. D. Griffiths, “Advances in Bistatic Radar”, Scitech Publishing Inc., 2007 12. R. Welsh, D. Andre, M. Finnis, “Volumetric interferometry for sparse 3D synthetic aperture radar with bistatic geometries”, Electronics Letters, 59 (12) Article No. e12851., 2023