Object detection from aerial and satellite remote sensing images has been an active research topic over the past decade. Thanks to the increase in computational resources and data availability, deep learning-based object detection methods have achieved numerous successes in computer vision, and more recently in remote sensing. However, the ability of current detectors to deal with (very) small objects still remains limited. In particular, the fast detection of small objects from a large observed scene is still an open question. In this work, we address this challenge and introduce an enhanced one-stage deep learning-based detection model, called You Only Look Once (YOLO)-fine, which is based on the structure of YOLOv3. Our detector is designed to be capable of detecting small objects with high accuracy and high speed, allowing further real-time applications within operational contexts. We also investigate its robustness to the appearance of new backgrounds in the validation set, thus tackling the issue of domain adaptation that is critical in remote sensing. Experimental studies that were conducted on both aerial and satellite benchmark datasets show some significant improvement of YOLO-fine as compared to other state-of-the art object detectors.

Convolutional neural networks (CNNs) have rapidly become the state-of-the-art models for image classification applications. They usually require large groundtruthed datasets for training. Here, we address object identification and recognition in the wild for infrared (IR) imaging in defense applications, where no such large-scale dataset is available. With a focus on robustness issues, especially viewpoint invariance, we introduce a compact and fully convolutional CNN architecture with global average pooling. We show that this model trained from realistic simulation datasets reaches a state-of-the-art performance compared with other CNNs with no data augmentation and fine-tuning steps. We also demonstrate a significant improvement in the robustness to viewpoint changes with respect to an operational support vector machine (SVM)-based scheme.

The classification of radar emitters in a naval electronic warfare context is a challenging problem. Furthermore, over the last few years, new technologies and the reduction of the fabrication costs have led to an increase in the number of radar. Thus, electronic warfare systems need to improve their capabilities of radar identification. The common operational methods, which consist in clustering pulses with similar parameters, are no longer able to solve the classification task efficiently. The main challenges of this data clustering of unknown radar emitters from passive measurements are the high dimension of the received radar pulse samples, the small sample group size, and closely located radar pulse clusters. In this paper, we propose to model the instantaneous frequency law (IFL) of the recorded pulses using B´ezier curves. The spatial distribution of the control points of the B´ezier curves are used as new features for the classification process based on the common density-based spatial clustering of applications with noise (DBSCAN) algorithm. Real data recorded near Brest, France, are used to highlight the interest in using the proposed approach. We also show the potential of the approach for specific emitter identification.

This paper deals with the design of sensor arrays in the context involving the localization of a few acoustic sources. Sparse approximation is known to be effective to find the source locations, but it depends on different array characteristics, such as the number of sensors and the array geometry. The present paper tackles this array design problem under the form of a sequential sensor selection procedure. The proposed method alternates between two steps. One step involves a source localization estimator, given a current set of measurement points, to obtain the estimation variance. Then, the other step selects the new point where a future measurement will maximally decrease the variance from the previous step. As such, the procedure can be applied online. Both numerical and experimental studies are conducted in an indoor nearfield configuration. Results show that the proposed approach performs better than offline state-of-the-art methods, and the presented empirical study reveals a better robustness to the model mismatches originating from the room reflections.