Conference Agenda

.

The growing demand for enhanced situational awareness in security and surveillance applications has driven significant advancements in Synthetic Aperture Radar (SAR) technologies. In this context, developing and optimizing SAR processing algorithms, especially those that use angular diversity through distributed SAR systems, has become crucial for improving detection, classification, and interpretation capabilities. The Next Generation Processing Methods for Security Applications (NGSECAPP) study in the frame of the ESA-ESRIN contract has concentrated on creating, validating, and evaluating advanced SAR processing algorithms designed to exploit the unique advantages of angular diversity found in distributed radar configurations. Several use cases across four main categories were implemented during this study, delivering a wide range of capabilities. A brief description of the use cases is provided below.

1) GMTI/MMTI in pursuit monostatic mode and with the staring spotlight mode:

Detecting moving targets typically requires at least two independent antennas spaced along the track. Using two satellites in a pursuit monostatic configuration enhances detection by reducing clutter and allowing precise estimation of a target’s position, velocity, and direction. This enables reliable tracking and prediction, which is particularly useful for maritime monitoring.

When a moving target is captured by a SAR sensor with very high resolution during its integration time, it causes blurring in the focused image. By observing the moving target from two different look angles, azimuth velocity can be estimated.

2. High-spatial resolution with continuous coverage:

While advanced digital beamforming SAR systems improve resolution and coverage but are costly and slow to evolve, NewSpace SAR constellations offer a cost-effective alternative by combining Doppler spectrum segments from multiple satellites to produce higher-resolution images through post-processing.

3. RFI mitigation with multi-channel systems:

This use case addresses the detection, identification, and localization of radio frequency interference (RFI) in multichannel SAR systems by exploiting coherence differences between receive channels on a pulse-by-pulse basis. By analyzing correlation in both time and frequency domains, the approach enables reliable RFI detection, suppression through techniques such as notch filtering, and estimation of the interference’s azimuth direction of arrival.

4a) Topography estimation with spotlight data:

Very high-resolution SAR imaging improves Earth observation but creates challenges in accurate 3D object localization due to elevation-induced defocusing and atmospheric effects, which single-image methods struggle to resolve. Using stereoscopic image pairs, especially from simultaneously acquired bistatic systems, enables better target recognition, separation of height and atmospheric defocusing effects, and more accurate absolute height estimation.

Synthetic Aperture Radar (SAR) imagery, captured by constellations such as Sentinel-1, SAOCOM, NISAR and Capella, is frequently affected by Radio Frequency Interferences (RFI) originating from terrestrial systems.

Traditionally discarded as image-degrading noise, these interferences hold significant analytical potential. The ongoing RUMORE (Retrieving Useful Measures frOm Radar intErferences) demonstration project, co-funded by the European Space Agency (ESA), is developing a methodology to extract precise Signal Intelligence (SIGINT) directly from RAW SAR data.
RUMORE operates directly on the uncompressed RAW data allowing the system to detect and isolate individual interference pulses to calculate precise Pulse Descriptor Words (PDWs), including centre frequency, bandwidth, pulse duration, and Pulse Repetition Frequency —a level of quantitative characterization impossible to achieve with other approaches based on focused data. These extracted parameters act as unique system fingerprint. A primary objective of the system is to systematically process both new acquisitions and extensive historical archives to populate a large-scale intelligence database with these fingerprints and their associated temporal activity. By matching these PDWs against known SIGINT databases, the system aims to classify emitter types and behaviours. Furthermore, RUMORE generates advanced analytical products such as emitter geolocation, temporal tracking, and Pattern-of-Life.

A validation campaign for the core algorithms has been completed with highly promising results.

Using reference emitters with public technical parameters, the validation demonstrated exceptional precision in signal characterization. The measured PDWs accurately matched expected official specifications for carrier frequency, bandwidth, and timing. Additionally, by combining multiple SAR acquisitions, the system successfully estimated the geographic positions of ground transmitters with a high degree of accuracy. This completed validation confirms that RUMORE’s outputs are highly reliable for populating large-scale intelligence databases with historical data. Ultimately, this ongoing project provides an opportunistic framework capable of natively fusing Image Intelligence (IMINT) and SIGINT from existing
SAR platforms.

Persistent wide-area surveillance requires change-detection tools that are both scalable and interpretable, especially in security contexts where analysts must rapidly discriminate genuine activity from artifacts. Sentinel-1 SAR time series provide all-weather, day–night coverage that is highly attractive for maritime domain awareness and anomaly screening. However, even after recent processing filtering methods, residual radio-frequency interference (RFI) may persist at low amplitude or intermittently, particularly in coastal transition zones where terrestrial emitters overlap maritime scenes.

We present a polarimetric optimization of the REACTIV change-visualization framework, designed to enhance the detection and separation of ultra-weak anomalies using dual-polarized Sentinel-1 amplitude time series (VV/VH) in linear scale. The method relies on a Multivariate Coefficient of Variation (MCV) formulation: a temporal covariance matrix is estimated in the joint polarimetric space and its eigenstructure yields two theoretical bounds, γ_−∞ and γ_∞. The lower bound γ_−∞ selectively emphasizes decorrelated or polarization-dependent temporal variations—typical of weak, residual RFI—whereas γ_∞ highlights coherent changes affecting both channels, as expected from legitimate targets or strong scatterers. A compact false-color composite (R=γ_−∞ , G=γ_∞, B=max(CV_VV,CV_VH) ) provides an analyst-friendly rendering that highlights meaningful activity without requiring complex modeling.
We demonstrate the approach in the Qatar coastal/offshore region (2024), a challenging security-relevant setting characterized by dense vessel traffic and intermittent low-level RFI. While single-channel CV maps reveal some interference, several traces appear in only one polarization; γ_−∞ further reveals ultra-weak structures that remain visually negligible in standard products, whereas γ_∞ preserves coherent maritime signatures. The method also helps disentangle mixed pixels where vessels and RFI coexist at different times. Implemented in a cloud-scalable workflow (e.g., Google Earth Engine), polarimetric REACTIV supports rapid, large-scale pre-screening for security operations, quality control, and downstream alerting pipelines.

Synthetic Aperture Radar (SAR) observations are increasingly susceptible to Radio Frequency Interference (RFI), which masks the useful target backscatter and significantly degrades the radiometric quality of the resulting imagery. The widespread proliferation of terrestrial electronic devices, including 5th generation (5G) mobile base stations and other high-power emitters, often introduces a relatively high Interference-to-Noise Ratio (INR) within the radar spectrum. As demand for very high-resolution imagery necessitates the need for wider signal bandwidths, the likelihood of spectral overlap with these external sources increases, leading to a higher prevalence of observed interference. Such artifacts manifest as high energy features in the frequency domain of Level-1 data, ultimately reducing the reliability of critical SAR applications.

To address this challenge, this paper presents an automated pipeline designed to detect and mitigate narrowband RFI signals by analyzing signal energy distributions within the spectral domain. The detection phase utilizes a U-Net architecture trained to perform semantic segmentation of complex RFI patterns against the background radar signal. By leveraging the spatial-spectral features of the data, the network achieves robust isolation of interference patterns. For the mitigation stage, a targeted notch filter is applied to the identified spectral bins, effectively suppressing the interfering energy while preserving the surrounding signal integrity. The proposed methodology is validated using StripMap mode data from the Airbus Radar Constellation. Experimental results demonstrate that the proposed solution reduces RFI-induced artifacts and restores image quality, offering an effective means of maintaining data fidelity in congested narrowband RF environments.

Synthetic Apertures Radar (SAR) technology has a great impact on Earth observation, thanks to its capability on acquiring high resolution images all-weather/day-night. Currently Its importance in the security and defense domain is steadily increasing, as SAR imagery provides valuable information for monitoring areas of interest and supporting geospatial intelligence and data-driven decision making. Moreover, compared with optical sensors, SAR systems are generally less susceptible to purely visual inflatable decoys, since radar backscatter depends on the electromagnetic and dielectric properties of observed targets. From here it arises the need of finding alternative techniques able to generate effective camouflage. In this work an active deception technique that allows to reproduce phony targets is explored. Specifically, the technique involves an active ECM, that, retransmitting the recorded SAR signal opportunely delayed and rephased, is able to reconstruct the odograph of a fictious target at a certain distance in slant range and in azimuth from it. System parameters requirements for the generation of the synthetic target signature are derived in different approximations of the point-target odograph, comparing their impact on the reconstructed signal over different image resolutions. The proposed models are tested through simulations performed in both medium-resolution (Stripmap) and high-resolution (Spotlight) acquisition modes using the end-to-end SAR simulator developed by Aresys. The ECM device is modeled like a radar with its own operation cycle, reception and transmission channel and the deception technique has been implemented as a C++ plugin solution of the Aresys simulator. It is worth to highlight that the analysis carried on considers spaceborne SAR systems, however the mathematical models used can be extended also to different platforms like drones or airplanes.

When multiple Synthetic Aperture Radar (SAR) receivers fly in close formation and observe the same scene, a Distributed SAR (DSAR) system is realized. This configuration enables the generation of a high-quality SAR image by coherently combining the lower-quality data collected by each receiver in the cluster. In the context of future spaceborne SAR missions, this capability is particularly attractive, as it allows high-quality imaging using constellations of smaller and less expensive satellites—such as CubeSats—rather than relying on large and costly platforms. This is the framework of the RODiO mission, funded by the Italian Space Agency (ASI) and currently undergoing Phase B. The system consists of four 16U CubeSats flying in close formation, with an along-track separation of tens of kilometers from the PLATiNO-1 satellite. Each CubeSat is equipped with a receive-only X-band antenna that collects bistatic echoes, using PLATiNO-1 as an illuminator of opportunity. In this way, RODiO aims to demonstrate the DSAR concept in orbit.

Within RODiO, the DSAR approach is exploited to reduce azimuth ambiguities and improve the Signal-to-Noise Ratio (SNR) compared to data acquired by a single platform. The baseline processing strategy consists in a combination of time-domain back-projection, suitable for handling the large-baseline bistatic geometry, with null-steering beamforming, which suppresses forward and backward azimuth ambiguities. The performance of the method is thoroughly analyzed considering both formation-flying aspects and imaging performance. The results confirm that the desired improvements in image quality can be effectively achieved in a wide range of working conditions. Nonetheless, very specific geometries exist, in terms of spatial separation among the receivers, in which performance is degraded because the beamforming problem becomes ill-conditioned. For the sake of keeping image quality even under those conditions, an alternative algorithm based on Wiener filter has been also selected and tested.

Here are presented several clock synchronisation and trajectory update techniques used for conducting bistatic SAR imaging experiments with airborne receiver and either cooperative airborne or non-cooperative space-borne transmitter.

Following an operational acquisition campaign for a bistatic database production, technique for radiometric calibration of bistatic SAR have been explored and the limits of the several autofocus and/or clock synchronisation approaches have been assessed.

With the non-cooperative space-borne transmitter (TerraSAR), due to its large antenna footprint, the receiving aircraft was directly illuminated allowing using the direct signal for clock synchronisation. Direct signal synchronisation was demonstrated either from a dedicated antenna, or from a mere leakage in the ground looking antenna (one-way propagation implies the direct signal is order of magnitude stronger at the aircraft than the ground scattered signal).

With cooperative aircraft transmitter, GPS clock disciplining proved a efficient synchronisation method for UHF band imaging.
At X band, however, this method was not accurate enough as GNSS is based on L-band signals.

Direct signal synchronisation proved difficult, if not impossible, at X band because of the narrow transmitter beam resulting the receiver being often out of the transmitter main-lobe, thus the direct signal was very low and plagued with multiple path disturbances.

Indirect signal through an active bistatic transponder was tested, and used when available, but this technique is only relevant to experimental operations as the active bistatic transponder is a complicated hardware which must track both aircraft and should be installed on ground within the transmitter antenna footprint.

The bulk of the operational acquisition campaign was synchronised using frame-drift auto-registration (autofocus but using a previous monostatic image as ground reference instead of matching successive images from the bistatic sensor) which proved robust and efficient.

For the radiometric calibration of the bistatic SAR system, the active bistatic transponder proved inaccurate because its synthetic RCS is very sensitive to minute tracking errors of both aircraft.

Using a quasi-monostatic calibration acquisition, i.e. an acquisition with transmitter and receiver aligned to target area and matching the sigma-nul of extended homogeneous area such as grass or forest area was accurate (provided the monostatic SAR image is available and calibrated) provided the angle between the antennae polarisation axes are carefully taken into account. However, the capability of being positioned in quasi-monostatic configuration is restrictive.

The stablest calibration technique tested was using a dihedral reflector from which a strong specular echo could be measured even for large bistatic angle and from which a bistatic RCS model is available. (top-hat and corner reflectors proved less usable calibrating targets).