Methods SAR

Based on an omnibus likelihood ratio test statistic for the equality of several variance-covariance matrices following the complex Wishart distribution with an associated $p$-value and a factorization of  this test statistic, change analysis in a short time series of multilook, polarimetric SAR data in the covariance matrix representation is carried out. The omnibus test statistic and its factorization detect if and when change(s) occur. The technique is demonstrated on airborne EMISAR C-band data but may be applied to Sentinel-1, Cosmo-SkyMed, TerraSAR-X, ALOS and RadarSat-2 data also. For ESA's Living Planet Symposium we intend to apply the methods sketched here on a series of Sentinel-1 data.

The two ESA Sentinel 1 and Sentinel 2 Satellite constellations will offer the opportunity of frequent and regular acquisitions all around the world in different spectral ranges. The measure of the interaction between the incoming active or passive radiation at the earth surface in different wave domains will intercept different physical land processes. Their combination will offer an unprecedented possibility to monitor land cover changes and to generate products useful for land management and emergency response.

In the framework of the ESA SEOM program, the "LAND COVER CHANGE DETECTION AND MONITORING METHODOLOGIES BASED ON THE COMBINED USE OF S1 AND S2 FOR NATURAL RESSOURCES AND HAZARD MANAGEMENT" project proposes a new approach for semi-automatic and probabilistic  mapping of land cover changes. this new method will combine the complementary capacities of optical and SAR data. The method will be tuned to track abrupt changes due to natural hazards such as landslides and floods, changes in land cover that influence natural hazard occurrence, like snow cover changes and forest changes, and changes in agriculture.

The framework consists of: (i) a multi-sensor training library of change signatures trapped by the Sentinels in a set-up phase and caused by landslides, floods, snow cover, deforestation and agricultural operations, and (ii) classifiers that use the elements of the library to automatically recognize, identify, and map new changes. The library represents a knowledge system which takes advantages from the combination of S1 & S2 data and that can be extended to other classes of changes.

Here the main goal of the project, the conceptual idea, the selected approach, and key applications will be presented.

Our research in exploiting the inherent ranging capabilities of Synthetic Aperture Radar (SAR) has demonstrated range observations at the one centimeter accuracy level for passive radar reflectors, given that the radar observations are processed by a geometrically accurate SAR processor. Moreover, external disturbances caused by atmospheric path delay (troposphere, ionosphere) and geodynamic displacements (solid earth tides, ocean loading, ...) have to be taken into account. The correction approach follows the concept used in well established geodetic techniques like Global Navigation Satellite System (GNSS) or radar altimetry. By making use of these geodetic SAR observations, direct 3D positioning of specifically designed passive objects (corner reflectors, octahedrons) or active transponders becomes attainable.

For TerraSAR-X, the 3D positioning capability has been tested with different passive reflectors and enables an accuracy at the centimeter level. Regarding Sentinel-1, the ranging results reported for the geometrical calibration campaign show encouraging results of approximately 6 cm, and further refinements are currently under investigation. In addition, the feasibility of active transponders as radar targets was successfully demonstrated, making Sentinel-1 another candidate for geodetic SAR applications.

The presentation discusses the 3D SAR method in the context of applications, which require an accurate monitoring of the ellipsoidal heights in order to link them to the global geometric reference frame. We give an overview on our SAR methods and state the potential for positioning and height monitoring with TerraSAR-X and Sentinel-1 on the base of selected study regions.

InSAR kinematic time series, describing the relative motion of virtual or real scatterers, suffer from a rank defect due to the lack of a proper datum definition. Consequently, the elevation or cross range coordinate is only known relative to some chosen reference, and the kinematic time series of each measurement location is dependent on the kinematic behavior of the reference point(s). This relativity affects all measurement points in terms of their position estimation as well as their kinematic behavior, which hampers the correct interpretation of the data.

Here we discuss the value of using low-cost transponders as reference points, by connecting these points to GNSS reference station antennas. This procedure fixes the InSAR scatterers to a Terrestrial Reference Frame, and allows for datum connection, quality control, and improved interpretation. We have tested the procedure for Radarsat-2 and Sentinel-1 time series, and report on the attainable quality levels.

.

Polarimetric Synthetic Aperture Radar (PolSAR) data are a very useful source of information about the earth surfaces for various applications such as land-cover classification, change detection or estimation of targets’ biophysical parameters. This wide range of applications is mainly a result of SAR environmental independencies. SAR data can be practically obtained in all weather conditions through day and night. The advantages of full polarimetry over common polarimetry modes are well-recognized.

Besides the fact that full polarized data usually have higher information content rather than any other polarization status, it also has some limitations e.g. the reduced swath width and the higher pulse repetition frequency (PRF) compared to dual polarization. In order to address a solution for these limitations, recently the concept of compact polarimetry has been introduced as an emerging idea, particularly for SAR remote sensing sensors. In this observation mode, either a linear polarized wave, oriented at 45° (л/4 mode), or a circular wave (л/2 mode) is transmitted, and then two linear polarized waves (H and V) backscattered form targets, are simultaneously received. Compact polarimetry can be considered as a dual polarimetry mode based on transmitting a combination of two linear polarized waves (H and V). In spite of common dual polarimetry modes, full polarimetry information can be reconstructed based on a few assumptions from compact polarimetry data. Despite all the invaluable advantages of full polarimetric data, in some applications where the symmetry properties assumption is valid (e.g. monitoring forests, crops, soil moisture, etc.), this technique comes with a significant drawback which is redundancy. However, the same required accuracy can be achieved with a lower cost through compact polarimetry.

In this study, the main objective is the evaluation of land-cover and crop classification potential of full and compact PolSAR observations. To this end, two data sets are used: first, the simulated compact polarimetry data from UAVSAR fully polarimetric images, acquired over an agricultural area in Winnipeg, Canada on 10^th July 2012; and second, the RISAT-1 compact (л/2 mode) and full polarimetry data acquired on 7^th Mar 2013 over India. The former data is intended to be used for a crop classification which suits the captured images from agricultural area, and the latter is used for a land-cover classification. For each data set, after applying calibration and multilook processing, full polarimetry covariance matrix is reconstructed from the compact polarimetry data. Then supervised Support Vector Machines (SVM) classification is applied to both polarimetric features extracted from full polarimetric and reconstructed full polarimetry covariance matrixes. The quantitative analysis has been done to assess the accuracy of land-cover map.