Many PSInSAR algorithms are based on the phase difference of connected PSs, which are usually spatially integrated to recover the deformation gradients with respect to a common reference point. In the presence of noise this can lead to errors propagating throughout the whole PS network. In our work we adapt the concept of connectivity originally proposed in the DInSAR context as a way to estimate the quality of the phase integration for each PS. We simulate PS networks under a variety of noise conditions, and find that connectivity is consistently correlated with integration errors, and provides complementary information compared to local quality parameters such as temporal coherence. We discuss the use of connectivity to optimize the selection of spatial references, and to discard measurements, which are more likely to be affected by integration errors. Finally, we validate our methodology on a real data set from TerraSAR-X covering the greater Copenhagen area.

This article presents a methodology for localizing radio frequency interference (RFI) signals in Synthetic Aperture Radar (SAR) images acquired from Sentinel-1 SAR satellites. RFI are caused by on-ground radars, and their detection and localization thus provide valuable information for decision makers. In this study, an unsupervised deep learning model based on a Convolutional Autoencoder is used to detect and localize RFI signals in SAR images. The CAE reconstructs the SAR images, excluding RFI signals and other large-scale anomalies. Anomalies are detected by comparing the original images with their reconstructions, and a secondary classification scheme is used to identify RFI signals among the detected anomalies. Results show that the proposed method detects and localizes RFI signals, even in complex regions. The automatic localization of RFI signals in SAR images can enhance various applications such as maritime domain awareness and border surveillance.

Synthetic Aperture Radar (SAR) satellite images are used increasingly more for Earth observation. While SAR images are useable in most conditions, they occasionally experience image degradation due to interfering signals from external radars, called Radio Frequency Interference (RFI). RFI affected images are often discarded in further analysis or pre-processed to remove the RFI. However, few on-ground radars can cause RFI in SAR images and such information can thus increase domain awareness greatly over both land and sea, where,e.g., localizing and characterizing RFI signals in the ocean could help classify otherwise overlooked ships. The aim of the current study is to detect and localize RFI signals automatically in Sentinel-1 level-1 images and further characterize the on-ground radar. The spatial structure of RFI signals vary greatly. A convolutional autoencoder was therefore developed to reconstruct RFI-free Sentinel-1 images. Conversely, RFI-affected images could not be well reconstructed. Anomalous heatmaps were then developed to automatically detect and localize RFI anomalies in the images under varying environmental and geographical conditions whereafter the external radar characteristics were extracted manually from Sentinel-1 level-0 data. We could consequently classify and localize RFI signals believed to originate from both stationary radars and ship-borne radars. We further argue that the calculated ship-borne radar characteristics correspond to those of air-surveillance radars. Empirically, the method showed better detection results than those of previous studies. Our study shows that more information can be extracted from certain detected objects, such as ships, from SAR images.

The application of Sentinel-1 interferometry to ice velocity measurements has until recently been limited by the significant horizontal scene motion associated with ice flow, which causes phase discontinuities (and associated unwrapping problems) at burst boundaries in Terrain Observation by Progressive Scans (TOPS) interferograms. Coregistering with a multiyear averaged external velocity mosaic based on offset-tracking can account for the bulk of the ice motion, but residual discontinuities sometimes remain, for example, due to seasonal variations in the ice velocity, or due to error sources such as azimuth shifts caused by ionospheric propagation. The presented method extends the external velocity coregistration with a local, spatially varying, coregistration in the burst overlap regions. This is based on the extended spectral diversity principle, which can only be applied in the overlap regions, but offers superior accuracy and resolution compared with traditional coregistration methods. The method considerably reduces phase discontinuities at burst boundaries, and potential new phase discontinuities at the overlap region edges are suppressed by an azimuth tapering of the applied coregistration shifts. An example scene is presented, and the phase discontinuities before and after application of the method are evaluated. The method is seen to remove phase discontinuities, with no adverse effects.

Differential synthetic aperture radar interferometry (DInSAR) allows for retrieval of ice velocity measurements of high resolution and accuracy. One of the main error sources in DInSAR is the phase unwrapping procedure. Unwrapping errors may be caused by several processes, including shear stresses associated with large motion gradients, which lead to loss of interferometric coherence. In many cases, unwrapping errors reach magnitudes corresponding to velocities of tens or even hundreds of meters per year. Traditional DInSAR implementations include pixel masking based on coherence thresholding; however, such a masking is not always sufficient. Consequently, the state-of-the-art for ice velocity retrievals involves either manual inspection of individual measurements or simply discarding measurements in regions where ice flow exceeds a predefined threshold. Here, we instead apply a masking based on thresholding of a pixel connectivity estimate with respect to a reference point, which aims to detect unwrapping errors based only on the estimated coherence pattern. The method is tested on both simulated and real data Sentinel-1 data from the Greenland Ice Sheet and effectively detects the majority of unwrapping errors (recall of 0.84 for the best performing threshold), although with a relatively low precision (0.52 for the best performing threshold). Importantly, higher magnitude unwrapping errors are associated with lower connectivity values, meaning that undetected errors have a significantly lower magnitude (median of 1.7 m/y, corresponding to a single phase cycle, compared with 40.5 m/y with no masking).

Data-driven classification algorithms have proved to do well for automatic target recognition (ATR) in synthetic aperture radar (SAR) data. Collecting data sets suitable for these algorithms is a challenge in itself as it is difficult and expensive. Due to the lack of labeled data sets with real SAR images of sufficient size, simulated data play a big role in SAR ATR development, but the transferability of knowledge learned on simulated data to real data remains to be studied further. In this letter, we show the first study of Transfer Learning between a simulated data set and a set of real SAR images. The simulated data set is obtained by adding a simulated object radar reflectivity to a terrain model of individual point scatters, prior to focusing. Our results show that a Convolutional Neural Network (Convnet) pretrained on simulated data has a great advantage over a Convnet trained only on real data, especially when real data are sparse. The advantages of pretraining the models on simulated data show both in terms of faster convergence during the training phase and on the end accuracy when benchmarked on the Moving and Stationary Target Acquisition and Recognition data set. These results encourage SAR ATR development to continue the improvement of simulated data sets of greater size and complex scenarios in order to build robust algorithms for real life SAR ATR applications.

Glacier and ice sheet dynamics are currently mapped with X-, C-, and L-band SAR. With the prospect of a P-band SAR, Biomass, to be launched within the next decade it is interesting to look into the potential of P-band for ice velocity mapping. In this paper first results are presented. Airborne P-band SAR data have been acquired in Greenland, and both offset tracking and DInSAR have been applied to the full resolution data as well as to data degraded to the resolution of Biomass. Generally, ice velocity maps are successfully generated, but in the ablation zone, DInSAR fails in the melt season at both resolutions, while feature tracking fails at the coarser resolution.

In February 2011, the Polarimetric Airborne Radar Ice Sounder (POLARIS) was flown in Antarctica in order to assess the feasibility of a potential space-based radar ice sounding mission. The campaign has demonstrated that the basal return is detectable in areas with up to 3 km thick cold ice, in areas with up to several hundred meters thick warm shelf ice, and in areas with up to 700 m thick crevassed glacier ice. However, major gaps in the basal return are observed, presumably due to excessive absorption, scattering from ice inclusions in the firn, low basal reflectivity, and the masking effect of the surface clutter. Internal layers are observed down to depths exceeding 2 km. The polarimetric data show that the internal layers are strongly anisotropic at a ridge, where the ice flow is supposed to be highly unidirectional. In case of space-based ice sounding, the antenna pattern cannot offer sufficient surface clutter suppression, but improved clutter suppression has been demonstrated with novel multi-phase-center techniques.

SAR processing can be applied to ice sounder data to improve along-track resolution and clutter suppression. This paper presents a time-domain back-projection technique for SAR focusing of ice sounder data. With this technique, variations in flight track and ice surface slope can be accurately accommodated at the expense of computation time. The back-projection algorithm can be easily parallelized however, and can advantageously be implemented on a graphics processing unit (GPU). Results from using the back-projection algorithm on POLARIS ice sounder data from North Greenland shows that the quality of data is improved by the processing, and the performance of the GPU implementation allows for very fast focusing.

ESA has assigned the Technical University of Denmark to develop an airborne P-band ice sounding radar demonstrator. The intention is to obtain a better understanding of the electromagnetic properties of the Antarctic ice sheet at P-band and to test novel ice sounding techniques in preparation for a potential spaceborne ice sounding radar. The airborne system is a coherent, high-resolution and fully polarimetric radar. Aperture synthesis is applied in the along track direction and an experimental surface clutter suppression technique based on a multi-aperture antenna can be applied in the across track direction. In May 2008, a proof-of-concept campaign was organized in Greenland, where data were acquired over the ice sheet. The system proved capable of detecting the bedrock under 3 km thick ice and of mapping the internal ice layers down to a depth of at least 1.3 km. In this paper, the system concept is outlined and first results are presented.

This paper presents the top-level design of an airborne, P-band ice sounding radar under development at the Technical University of Denmark. The ice sounder is intended to provide more information on the electromagnetic properties of the Antarctic ice sheet at P-band. A secondary objective is to test new ice sounding techniques, e.g. polarimetry, synthetic aperture processing, and coherent clutter suppression. A system analysis involving ice scattering models confirms that it is feasible to detect the bedrock through 4 km of ice and to detect deep ice layers. The ice sounder design features a digital signal generator, a microstrip antenna array, a conventional RF-architecture with a central transmitter, four receivers, and internal calibration loops. In 2008 the first data acquisition campaign will take place in Greenland.

A novel ambiguity suppression technique is proposed. Range ambiguities in synthetic aperture radar (SAR) images are eliminated with an azimuth filter after having applied an azimuth phase modulation to the transmitted pulses and a corresponding demodulation to the received pulses. The technique excels by actually eliminating the ambiguities rather than just defocusing them as most other techniques do. This makes the proposed technique applicable to distributed targets. The range ambiguity suppression permits the pulse repetition frequency (PRF) to exceed the upper limit otherwise defined by the antenna elevation dimension. The fundamental antenna area constraint still applies, but the PRF can be chosen with more freedom. In addition to ambiguity suppression, potential applications include nadir return elimination and signal-to-noise ratio improvement