Björn Tings

First Comparison of Sentinel-1 and TerraSAR-X Data in the Framework of Maritime Targets Detection: South Italy Case

The Sentinel-1A is the first of two satellites that composes the Sentinel-1 radar mission. Both satellites operate a C-band synthetic aperture radar (SAR) system to give continuity to the European SAR program. SAR is a flexible sensor able to fulfil users/applications requirements in terms of resolution and coverage thanks to different operational modes and polarizations. With the in-orbit availability of very-high-resolution X-band SAR sensors, the Sentinel-1 satellites have been designed to achieve wide coverage at medium to high resolution. The interferometric wide swath (IWS) mode implemented with the terrain observation with progressive scan (TOPS) technique is the standard acquisition mode over European waters and land masses. IWS in dual-polarization (VV/VH) combination offers 250-km swath at 5 m × 20 m (range × azimuth) spatial resolution. These specifications are in line with the needs of the European Maritime and Security Agency (EMSA) for oil spill and ship detection applications included in the CleanSeaNet program. The main goals of this paper are: assessment of medium-to-high-resolution C-band Sentinel-1 data with very-high-resolution X-band TerraSAR-X data for maritime targets detection; synergetic use of multiplatforms satellite SAR data for target features extraction; evaluation of polarimetric target detectors for the available co-polarization and cross-polarization Sentinel-1A IWS VV/VH products. The objectives are achieved by means of real, almost coincident C-band and X-band SAR data acquired by Sentinel-1A and TerraSAR-X satellites over Gulf of Naples and Catania (South Italy). Furthermore, the obtained results are supported by recorded ground truth vessel reports via terrestrial automatic identification system (AIS) stations located in the area.

Ship Classification in TerraSAR-X Images With Convolutional Neural Networks

Synthetic aperture radar (SAR) is an important instrument for oceanographic observations, providing detailed information of oceans’ surface and artificial floating structures. Due to advances in SAR technology and deployment of new SAR satellites, an increasing amount of data is available, and the development of efficient classification systems based on deep learning is possible. A deep neural network has improved the state of the art in classification tasks of optical images, but its use in SAR classification problems has been less exploited. In this paper, a full workflow for SAR maritime targets detection and classification on TerraSAR-X high-resolution image is presented, and convolutional neural networks (CNNs) recently proposed in the literature are cross evaluated on a common data set composed of five maritime classes, namely, cargo, tanker, windmill, platform, and harbor structure. Based on experiments and tests, a multiple input resolution CNN model is proposed and its performance is evaluated. Our results indicate that CNNs are efficient models to perform maritime target classification in SAR images, and the combination of different input resolutions in the CNN model improves its ability to derive features, increasing the overall classification score.

Dynamically adapted ship parameter estimation using TerraSAR-X images

This article describes how the estimation of ship parameters from ship signatures on TerraSAR-X images can be adapted dynamically using combinatorial optimization and regression analysis. Research in the field of ship detection commonly addresses the improvement of processors with regard to accuracy performance of detection and parameter estimation. While most research implies beneficial improvements to the processors, the different techniques are rarely compared or combined. In this article the Monte Carlo combinatorial optimization (cross-entropy method) is used to evaluate the performance of improvements to parameter estimation and performance of combinations of these improvements. Then multiple linear regression analysis is applied to increase the accuracy of parameter estimation further. The underlying data set consists of TerraSAR-X Stripmap, ScanSAR, and ScanSAR Wide Multi Look Ground Range detected (MGD) images acquired over the North Sea and Baltic Sea with horizontal transmit, horizontal receive or vertical transmit, vertical receive polarization. Validation data are provided by the Automatic Identification System (AIS). The optimization algorithm assesses optimal parameter settings and appropriate combinations of techniques dedicated to this data set. The resulting processor provides a significantly higher accuracy of ship parameter estimation than the initial processor.

Comparison of Sentinel-1 and TerraSAR-X for ship detection

Sentinel-1A (S1-A) is the first of a pair of satellites operating a C-band Synthetic Aperture Radar (SAR) developed to give continuity to the European SAR programme. After 6 months commissioning phase from its lunch on April 3rd 2014, S1-A started to systematically deliver data to Copernicus Ocean, Land and Emergency services. Interferometric Wide Swath (IWS) mode in dual-polarization VV+VH is the standard acquisition mode used for the observation of the European marine environment. The objectives of this study are the comparison and synergetic use of the S1-A imagery in IWS mode with TerraSAR-X (TS-X) imagery in StripMap mode for SAR ship detection application. Exemplary dataset acquired over the English Channel during a controlled experiment is used to pursue the objective. Automatic Identification System ships messages received by terrestrial stations in the area are used as ground truth and to identify the SAR detections.

Comparison of ship wake detectability on C-band and X-band SAR

ABSTRACT This article describes how a detectability model can be trained in the form of a binary classifier from a data set of synthetic aperture radar (SAR) images of ship wakes, augmented by automatic identification system data. While detectability models for ship signatures exist, ship wake detectability models are only available for simulated data. In order to improve existing ship wake detection algorithms on SAR imagery, there is a need for building a data-driven detectability model which may provide useful a-priori information. A binary L2-regularized logistic regression classifier is trained for each investigated data subset. The dependency on the SAR working frequency is evaluated by analysing a large number of X- and C-band images. In the X-band, the probability of detecting a wake shows dependencies on vessel size and velocity as well as prevailing wind speed. In the C-band, these dependencies are maintained, but with a general reduction in the correlation. This fact led us to the conclusion that, for our data set, ship wakes are more easily imaged in the X-band rather than in the C-band. This is an important outcome, which is supported by a qualitative and quantitative analysis of a large data set of TerraSAR-X and two independent C-band sensors, specifically RADARSAT-2 and Sentinel-1.

Comparison of ship detectability between TerraSAR-X and Sentinel-1

In this paper, the detectability of ship signatures in Synthetic Aperture Radar (SAR) imagery acquired by the TerraSAR-X/TanDEM-X and Sentinel-1 is compared. The comparison takes into account different sensors acquisition parameters and environmental conditions on a large variety of ship size and types. In the first step, ocean targets are detected using the Near Real Time (NRT)-optimized Constant-False-Alarm-Rate (CFAR) algorithm. The optimizations include the ocean/land and false targets discrimination. In the second step, all detected targets are automatically matched in space and time with the recorded Automatic Identification System (AIS) messages. A manual cross-check is performed at the end of the assignments to have a clean SAR ship signature database. Additionally, the local wind field is retrieved from the SAR backscatter of the ocean surface surrounding the detected ships, by applying the Geophysical Model Functions (GMF) inversion XMOD2 for X-band data and CMOD5 for C-band data. Similarly, the local sea state conditions are calculated by the XWAVE and CWAVE empirical model functions. The final detectability model takes into account all SAR-based information, i.e. wind speed and sea state, as well as relevant SAR parameters, e.g. incidence angle. The overall probability of detection are derived for three ship size categories, i.e. small, medium and large, adopting an L2-regularized Logistic Regression classifier trained on detected and nondetected ship samples.