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.

Synthetic Aperture Radar (SAR) provides detailed information of Ocean's surface and man-made floating structures. Advances in SAR technology and deployment of new SAR satellites have contributed to an increasing number of remote sensing data available. Handle this large amount of data with human operators is infeasible. Therefore, the use of automated tools to process remote sensing images, identify regions of interest, and select relevant information are needed. The use of neural networks to solve SAR image classification problems is well known. A typical architecture consists of a shallow feed-forward neural network with an input layer, a hidden layer, and an output layer. This type of neural network, combined with back-propagation and a gradient-based training algorithm, is able to solve complex problems in SAR image analysis. However, this architecture is unable to take advantage of unlabeled data during its training process, and in many cases the input features need to be carefully tuned in order to reduce the overall network complexity. This paper proposes the application of Deep Neural Networks (DNN) to perform oceanographic-object classification.

Different metocean conditions have an impact on the detectability of ship signatures on Synthetic Aperture Radar (SAR) images. During the EMSec Project algorithms for retrieval of wind and sea state fields from TerraSAR-X data have been developed in conjunction with a near real-time-capable constant false alarm rate ship detection processor. This paper presents a new model connecting these three information extraction systems into a ship detectability model by setting the probability of detection in dependency to the four parameters: Wind speed, significant wave height, incidence angle and ship length. The model is based on a binary L2-regularized logistic regression classifier trained on a large dataset of X-band SAR ship samples, which are identified using Automatic Identification System messages co-located automatically in space and time and further checked manually to avoid possible mismatches. Results are compared to the state-of-the-art simulation-based ship detectability model available in literature. For the first time it has been possible to evaluate not only qualitatively but also quantitatively the effects of acquisition geometry and metocean conditions for the different image resolution classes obtainable with the high-flexible SAR sensor on-board the TerraSAR-X satellite.

A dynamic swarm formation based in artificial potential fields is presented and extended to 3D environments. A dynamic technic to adjust the swarm configuration when avoiding obstacles is also proposed. Using potential fields in conjunction with a global planner like the A* algorithm allows the navigation through an environment in presence of obstacles. The global planner finds an obstacle free path using a discrete representation of the world and this path is used to guide the center of the swarm formation. Potential field is used to control the swarm formation and to avoid obstacles locally when travelling the planned path. Simulation results are presented for swarm of particles of different sizes in 2D and 3D worlds.

A growing interest in global maritime surveillance has motivated the development of new methods and sensors to overcome the limitations of coastal-based systems coverage. Among the available state of the art tools, the Synthetic Aperture Radar (SAR) stands out as a powerful sensor for oceanographic observation and maritime surveillance, due to its capability of wide swath coverage and the ability to operate in all weather conditions. This paper examines the minimum ship size detectable by the Constant False Alarm Rate (CFAR) algorithm over different TerraSAR-X product modes, using simulation-based analyses. These analyses allow a comparison of the expected CFAR ship detection performance for different TerraSAR-X products, over different sea state conditions.