Conference Agenda

ONERA, the French aerospace lab, is one of the leading organizations in experimental research on aeronautic topics. Works on airborne synthetic aperture radars began almost fourty years ago with the first airborne demonstrator named RAMSES, on board a Transall C160 carrier aircraft. Later ONERA developed on 2006-2009 two new SAR platforms flying on board a STEMME S10-VTX and a Falcon 20 jet respectively.

The first one, called BUSARD, allowed multiple airborne data collection campaigns for both civilian and defense applications. SAR sensors available on board BUSARD are X and Ka bands, with both polarimetric and interferometric capacities. Some experimentations have also been promoted with passive sensors at lower bands (L, C, S bands).

The second one, the well-known SETHI, participated to its first operational campaign on 2009, for ESA, in the frame of BIOMASS mission. It was the first TropiSAR campaign that collected a large SAR dataset over French Guyana. Then SETHI participated to various experimentations for defense and security applications, such as: High-resolution SAR imaging, coherent and incoherent change detections, GMTI measurements, very long range operations, circular imagery…

Our presentation will summarize the work carried out by SETHI and BUSARD and their contributions to airborne radar remote sensing research these last 20 years, for both defense and security applications.

Spaceborne synthetic aperture radar (SAR) represents a fundamental means to address the challenges of the next decades in the domains of Earth monitoring, as well as intelligence, surveillance and reconnaissance (ISR). Novel ambiguous SAR acquisition modes, recently demonstrated with TerraSAR-X, can circumvent the swath/resolution constraint without the need for digital beamforming. These modes are effective for dedicated applications, such as maritime surveillance and ground deformation monitoring, leading to a coverage increase by a factor of 20 with the same technology. SwitchSAR is a recently-developed SAR concept that enables a multi-fold increase of the sensitivity of a SAR system for the same average transmit power and is particularly attractive for Ka-band systems and topographic interferometric applications. Besides the generation of accurate digital elevation models, the coherent combination of SAR images, taken from different angles, helps unveil the three-dimensional structure of vegetation, ice, and dry soil through high-resolution tomograms that can be better interpreted if acquired using multiple-input multiple-output systems, as demonstrated using a ground-based radar. Distributed systems, based on clusters of small satellites flying in formation, enable simultaneous collection of such multi-angular images with a significant impact on numerous applications. An affordable and versatile approach for their demonstration is based on swarms of drones equipped with lightweight radar sensors, which DLR is implementing by building up a distributed drone-based SAR infrastructure. Bistatic and multistatic geometries with spaceborne transmitters and airborne receivers also play a fundamental role in ISR applications, where broadband communication satellites can be exploited as illuminators of opportunity. These advances are forerunners for the development of novel full-fledged and NewSpace spaceborne Earth observation missions that will offer remarkable societal benefits.

The aiRadar MMI-series multi-channel millimeter-wave radars have been deployed on small- and medium-sized drones as interferometric SAR / GMTI systems. There are several major differences in the operating regime of these low-flying drone-based mmWave systems when compared with existing airborne (and especially spaceborne) SAR sensors. Some of these differences have positive practical effects, such as low operational cost, reduced deployment complexity, increased flexibility in acquisition timing and geometry, and fine range resolution without large fractional bandwidth. Other differences cause problems for data processing, like the large variation of incidence angles across the imaging swath, magnified influence of topography on scene geometry, and extremely poor position and attitude control (especially relative to the short wavelength). Existing techniques for mitigating motion errors require complex, heavy, and expensive IMUs along with multiple iterations of computationally costly autofocus algorithms to estimate the chaotic motion of the drone and obtain well-focused images. We have developed navigation algorithms which leverage the multi-aperture nature of the aiRadar MMI sensors to estimate and compensate for the irregular motion of the platform using only consumer-grade IMUs and minimal autofocus. Results from initial deployments on two different drone platforms demonstrate the effectiveness of our multi-aperture radar-data-driven navigation approach. With this system we have achieved precise navigation of the aircraft, high-resolution focusing of SAR data, interferometry, polarimetry, GMTI, and multi-aspect imaging. Work is ongoing to develop real-time versions of our navigation and focusing software to enable onboard real-time navigation and image formation. Our precise navigation capability and the advanced radar techniques it permits make the aiRadar DroneSAR system useful both for advanced mmWave SAR imaging in its own right and for the simulation of other airborne and spaceborne radar systems.

SAR (synthetic aperture radar) provides high-resolution images using radar electromagnetic waves acquired from airborne or satellite platforms. SAR change detection is an important application since it allows to monitor designated sites. SAR change detection has been firstly developed using SAR imaging alone. However, SAR systems can operate at multiple frequencies and multiple channels. SAR change detection can benefit from additional degrees of freedom, allowing more diversity in the acquired data of an observed scene. Polarization is a common feature of SAR systems that allows one to generate SAR images for four different polarimetric channels. Including multiple polarizations improves SAR change detection performances, especially for false alarms regulation. Multi-temporal acquisition has also been extensively investigated for site monitoring. At last, SAR systems can acquire data from a range of frequency bands. A scene or objects show a variety of scattering properties according to the frequency of the SAR signal.

As more SAR systems are available, more applications using SAR images acquired from multiples frequencies have been proposed in these last 10 years. Electromagnetic scattering strongly depends on the used frequency. Structural and physical information from objects and scenes can be derived from their response in different frequencies. Multi-frequency SAR has been firstly applied to terrain and vegetation classification to improve performances. Bi-date change detection using SAR images acquired at different times in multiple frequencies has been proposed using a joint covariance matrix of the PolSAR multi-frequency data. This approach is studied and improved by introducing Kronecker product to combine frequency information into a general multi-modal framework. Change vector analysis is also presented and consists in using frequency bands to classify changes. More recently, a change detection framework based on multivariate statistics has been proposed to optimally combine the multi-frequency SAR images.

In this work, we propose to focus on the characterization of the changes between 2 multi-frequency SAR acquisitions using single or multiple polarizations. First, analysis is done on the scattering properties of the objects of the scene in different frequencies. Polarisation decompositions are also considered to give more insight on the multi-frequency scattering behaviours. Secondly, change detection characterization is considered using approach based on CVA (change vector analysis). Change detection is performed separately for each frequency in single or multi-polarizations. Analysis of detected changes obtained in different frequencies is then done by fusing them. The simplest way to apply CVA is to compute the ratio of the change detections between frequencies. This ratio gives a first CVA by measuring the amount of changes due to each frequency. Multi-class unsupervised classification is also considered as it permits the definition of several classes for CVA. Results have been obtained on an airborne dataset in L and X bands, full polarization, acquired by the ONERA sensor SETHI in the south of France with resolutions of 1m in L-band and 30cm in X-band.

.