Non Stationary Bistatic Synthetic Aperture Radar Processing: Assessment of Frequency Domain Processing from Simulated and Real Signals

Transcription

1 PIERS ONLINE, VOL. 5, NO. 2, Non Stationary Bistatic Synthetic Aperture Radar Processing: Assessment of Frequency Domain Processing from Simulated and Real Signals Hubert M. J. Cantalloube Office National d Études et Recherches Aérospatiales, France Abstract Bistatic synthetic aperture radar (SAR) imaging from very asymmetric configuration is a promising technique for both military and civilian issues. Indeed, an illuminating radar standing off at a safe distance may be combined with a low cost, possibly unmanned air vehicle using a passive radar receiver operating at closer range. Practical civilian application could be high resolution remote sensing of dangerous disaster areas (fire, chemical or radioactive hazard) with small unmanned aircrafts. Military application could be SAR imaging in the forward direction for a missile guidance without signalling the sensor by its transmitted radiation. However, such configurations are strongly non-stationary in the sense that the transmitter to receiver distance and relative orientation varies. This severely harden the task of frequency domain processing and especially its motion compensation. We tested frequency domain processing and motion compensation for both simulated and real signal for identical asymmetric configurations. The SAR processor may provide self-testing before image synthesis and forecast phase errors in the resulting image depending on terrain elevation features. Error maps provided may be used for illustrating the motion compensation and the frequency domain processing in a didactic way. Opportunistic air-to-air ISAR imaging (of the receiver plane) was successfully experimented, though bistatic imaging was mostly a failure due to local clock jitter. This crucial issue as well as the clock drifting issue will be addressed. 1. INTRODUCTION Following the first airborne bistatic campaign [1] conducted by the German DLR and the French ONERA in 2003, which successfully tested time invariant (same velocity on parallel tracks) configuration, we used the last two flights before decommissioning of our older airborne SAR system RAMSES (on October 21st and 25th 2008) for testing strongly time-varying configurations using as receiver our lightweight DRIVE system. Two representative configurations were selected with RAMSES (on board a Transall-C160, a large freighter aircraft) flying at 5000 feet/165 knots and illuminating at X-band from it right door and DRIVE as a receiver mounted on a pod under the right wind a small touring motor-glider (simulating an unmanned aircraft) flying at 3000 feet/85 knots with antenna squinted either 90 or 60 from the track. In the former case, the aircraft tracks are parallel and in the later case, their headings are separated by 30. In order to prepare this experiment, bistatic signal simulations for point-wise perfect reflectors were prepared using translated and rotated effective trajectories of the two aircrafts (recorded during independent monostatic acquisition campaigns) thus providing realistic motion compensation test data for our bistatic SAR processors. The first one used since 2003 is an accelerated temporal domain processor (a two-stage simpler and less efficient variant of the fast back projection algorithm [2]) the second one is an ω-k (frequency domain) processor derived from a motion compensated monostatic ω-k processor [3] and a bistatic processor prototype described in [4] and [5]. During this preparation stage, the idea of taking the opportunity to image the receiving aircraft itself and focusing its shadow was accepted for inclusion in the flight plan. 2. BISTATIC SAR PROCESSING ISSUES Our time-domain processor was already modified for bistatic configurations in 2003, and the time varying case adds but one complications: the integration time for a given range varies along track because the bistatic angle varies (the computation of it needed to be changed a first time in 2003 because in monostatic cases integration time is simply proportional to range, which is not the case in bistatic configurations with significantly separated parallel tracks). We did not implement this yet since our receiver system DRIVE can only record 16.5 second of signal, and this effect was not critical as the spectral envelop in azimuth is significantly degraded by the factoring of the back projection algorithm (even though it is limited to 2 stages in our implementation).

2 PIERS ONLINE, VOL. 5, NO. 2, The frequency-domain processor was simplified compared to the organisation described in [5]. Mainly, the motion-compensation parameters are computed in a preparation step and stored instead of being computed on the fly. The increase in required storage is traded against three advantages: First the motion/algorithmic compensation parameters are assessed before starting the time-consuming image synthesis, which allows for aborting the computation in case of insufficient values are detected. Second, the α-δ control loop during slow-time re-sampling (pre-integration) is replaced by a simple computation from the azimuth migration. Last, all the terrain altitude computations are pre-computed in the target image coordinate using a z-buffer type of approach, thus avoiding switching between solutions in overlay area which strongly disturbed the motion compensation mechanisms downstream in the processor. 3. REAL SIGNAL PROCESSING While processing of the simulated signal (as well as that of the 2003 time-invariant acquisitions) gave satisfying results, SAR processing of the just acquired signal produced images cluttered with azimuth replica of the expected landscape image (Fig. 1 and Fig. 2). Figure 1: From left to right: time-domain processing of simulated & real data, altitude, azimuth migration, 0th, 1st and 2nd order quadratic phase maps, resulting frequency-domain computed images. Figure 2: Monostatic stripmap image (left) obtained by frequency domain processing, monostatic flashlook image (right) time-domain processed and bistatic flashlook image (bottom) also time-domain processed. Illuminator trajectory is in red (light red not transmitting and dark red integration time for the flashlook images) and the trajectory of receiver only aircraft is in blue (light blue not receiving, dark blue integration time for the bistatic flashlook image).

3 PIERS ONLINE, VOL. 5, NO. 2, In order to explain the presence of azimuth image replica, we recorded, on the ground after the flights, the transmitted signal with the bistatic receiver (Fig. 3). Although the low frequency phase drift is comparable to the one dealt with during the DLR-ONERA campaign (the drift is about 5 time more slowly), there is an almost periodic higher (200 Hz) frequency component of yet undetermined origin. Figure 3: Phase drift measured by recording on the ground the transmitted signal of RAMSES with the DRIVE receiver. Phase history on 400 pulses (left) and the corresponding spectrum (right). Compensated bistatic flashlook image (bottom). 4. AIR TO AIR SAR IMAGING Knowing the receiver aircraft trajectory, its SAR imaging is performed by first registering the monostatic range profiles to its successive positions (Fig. 7) Optional low-pass filtering allows for important reduction of the data rate and increase of the target to clutter ratio which may be usefull in case range to target is less accurate than in our experiment. Range profile around the target centre are then processed by polar format algorithm using the radar trajectory in the coordinate space of the target aircraft. Unlike the classical surface-to-air or air-to-surface ISAR, air-to-air relative trajectory shows more elevation angle variation and uneven target rotation rate mostly because of target own rotations on the yaw axis (see on Fig. 4 the rotation rate inverts sometimes during the acquisition). Once these two phenomena are corrected (illustrated on simulated signal of 4 point-like reflectors on Fig. 4 and Fig. 5) it is possible to obtain coarse ISAR images of the real target, providing integration time is sufficiently low. Relative trajectory is then adjusted by the classical framedrift technique: The drift of successive images in the azimuth direction (measured by correlating detected images) is compensated for as radial velocity biases, then interpolated velocity correction is used to synthesize higher resolution images (Fig. 8). Figure 4: Apparent elevation (left) versus heading of the illuminating radar seem from the target aircraft during the true acquisition. Image (right) of a simulated point-wise echo without compensating for elevation angle fluctuations. Phase is colour-coded on the Fourier domain (right half of the windows) and shows in the former case fluctuations ruining azimuth (Doppler axis) resolution.

4 PIERS ONLINE, VOL. 5, NO. 2, Figure 5: Image of a simulated point-wise echo taking into account elevation angle fluctuations hence no phase error occurs. Without (left) compensation of the angular velocity fluctuations, uneven azimuth weighting increases the side-lobes. With (right) compensation, the image focusing is correct. Figure 6: Monostatic image of the last RAMSES acquisition (left) and the corresponding ground map (right). The true ground trajectory of the receiver aircraft is indicated as a fat blue line. Radar illumination is from the top of the illustration, moving from left (South) to right (North). Receiver trajectory is from the bottom-right (North-East) to the top-left (South-West). Further research will be to try focussing on the shadow of the target aircraft (as it is less altered by her attitude variations than her direct SAR image). Though it is relatively easy to compute the intersection with the digital terrain model (DTM) of the straight line joining illuminating and target aircrafts, hence the shadow position in range and Doppler (that of the ground at the shadow location), the SAR processor should be modified because unlike a real target, the Doppler is no more proportional to the first derivative of the range. Another axis of research is to try recovering attitude from tracking (may require ATI capability of imaging radar) through aircraft dynamics modelling. Figure 7: Range profiles centred on the target aircraft position (above: with normal rendering, below: with dynamic range increased to emphasize off clutter area. Top: genuine range profiles, bottom: low-pass filtered at 10% of the Doppler bandwidth). Times increases from left to right and range from top to bottom. Black area on the left corresponds to signal before the target enters the radar swath. The typical signature of the aircraft appears as horizontal trace at the bottom image.

5 PIERS ONLINE, VOL. 5, NO. 2, Figure 8: In flight SAR image (left) and optical image (middle) of the receiver aircraft, a Stemme touring motor glider, with a T shaped tail and the receiving radar under the right wing. Photo (right) on the ground of the receiver aircraft with the illuminating aircraft on the background. 5. CONCLUSIONS Though the experiment aim (SAR imaging in non stationary bistatic airborne configurations) was not attained, both the validation on simulated signal and the realisation of the first air-to-air radar imaging are promising results. The determination of the cause of failure (spectral purity of the receiver local oscillator, frequency multipliers, etc.) is also important as it will be a crucial point to check in the future bistatic system developments, especially in the receiver is a low-cost, possibly highly exposed to hazard or expendable system. ACKNOWLEDGMENT Author acknowledges funding and support from the DGA (French MoD research directorate) as well as kind cooperation from Istres test-range pilots and experimenters. REFERENCES 1. Dubois-Fernandez, P., et al., ONERA-DLR bistatic SAR campaign: Planning, data acquisition, and first analysis of bistatic scattering behaviour of natural and urban targets, IEE Radar Sonar Navigation, Vol. 153, No. 3, , Ulander, L., H. Hellsten, and G. Stenstrom, Synthetic aperture radar processing using fast factorized back-projection, IEEE Trans. Aerosp. Electron. Syst., Vol. 39, , Cantalloube, H.-M. J and P. Dubois-Fernandez, Airborne X-band SAR imaging with 10 cm resolution Technical challenge and preliminary results, Proceedings of IGARSS Conference, , Toulouse, France, July Giroux, V., et al., An omega-k algorithm for SAR bistatic systems, Proceedings of IGARSS Conference, , Seoul, Korea, July Cherniakov, M., Bistatic Radar: Emerging Technology, Wiley, New York, 2007.