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.

With a growing number of different satellite sensors, data fusion offers great potential in many applications. In this work, a convolutional neural network (CNN) architecture is presented for fusing Sentinel-1 synthetic aperture radar (SAR) imagery and the Advanced Microwave Scanning Radiometer 2 (AMSR2) data. The CNN is applied to the prediction of Arctic sea ice for marine navigation and as input to sea ice forecast models. This generic model is specifically well suited for fusing data sources where the ground resolutions of the sensors differ with orders of magnitude, here 35 km x 62 km (for AMSR2, 6.9 GHz) compared with the 93 m x 87 m (for sentinel-1 IW mode). In this work, two optimization approaches are compared using the categorical cross-entropy error function in the specific application of CNN training on sea ice charts. In the first approach, concentrations are thresholded to be encoded in a standard binary fashion, and in the second approach, concentrations are used as the target probability directly. The second method leads to a significant improvement in R² measured on the prediction of ice concentrations evaluated over the test set. The performance improves both in terms of robustness to noise and alignment with mean concentrations from ice analysts in the validation data, and an R² value of 0.89 is achieved over the independent test set. It can be concluded that CNNs are suitable for multisensor fusion even with sensors that differ in resolutions by large factors, such as in the case of Sentinel-1 SAR and AMSR2.

For disaster emergency response, timely information is critical and satellite data is a potential source for such information. High-resolution optical satellite images are often the most informative, but these are only available on cloud-free days. For some extreme weather disasters, like cyclones, access to cloud-free images is unlikely for days both before and after the main impact. In this situation, Synthetic Aperture Radar (SAR) data is a unique first source of information, as it works irrespective of weather and sunlight conditions. This paper shows, in the context of the cyclone Idai that hit Mozambique in March 2019, that Change Detection between pairs of SAR data is a perfect match with weather data, and therefore captures impact from the severe cyclone. For emergency operations, the filtering of Change Detections by external data on the location of houses prior to an event allows assessment of the impact on houses as opposed to impact on the surrounding natural environment. The free availability of SAR data from Sentinel-1, with further automated processing of it, means that this analysis is a cost-effective and quick potential first indication of cyclone destruction.

Convolutional Neural Networks (Convnets) have achieved good results in a range of computer vision tasks the recent years. Though given a lot of attention, visualizing the learned representations to interpret Convnets, still remains a challenging task. The high dimensionality of internal representations and the high abstractions of deep layers are the main challenges when visualizing Convnet functionality. We present in this paper a technique based on clustering internal Convnet representations with a Dirichlet Process Gaussian Mixture Model, for visualization of learned representations in Convnets. Our method copes with the high dimensionality of a Convnet by clustering representations across all nodes of each layer. We will discuss how this application is useful when considering transfer learning, i.e.\ transferring a model trained on one dataset to solve a task on a different one.

Spectral computed tomography (CT) is an emerging technology capable of providing high chemical specificity, which is crucial for many applications such as detecting threats in luggage. This type of application requires both fast and high-quality image reconstruction and is often based on sparse-view (few) projections. The conventional filtered back projection (FBP) method is fast but it produces low-quality images dominated by noise and artifacts in sparse-view CT. Iterative methods with, e.g., total variation regularizers can circumvent that but they are computationally expensive, as the computational load proportionally increases with the number of spectral channels. Instead, we propose an approach for fast reconstruction of sparse-view spectral CT data using a U-Net convolutional neural network architecture with multi-channel input and output. The network is trained to output high-quality CT images from FBP input image reconstructions. Our method is fast at run-time and because the internal convolutions are shared between the channels, the computational load increases only at the first and last layers, making it an efficient approach to process spectral data with a large number of channels. We have validated our approach using real CT scans. Our results show qualitatively and quantitatively that our approach outperforms the state-of-the-art iterative methods. Furthermore, the results indicate that the network can exploit the coupling of the channels to enhance the overall quality and robustness.