Understanding the decision-making process of deep learning networks is a key challenge which has rarely been investigated for Synthetic Aperture Radar (SAR) images. In this paper, a set of new analytical tools is proposed and applied to a Convolutional Neural Network (CNN) handling Automatic Target Recognition (ATR) on two SAR datasets containing military targets.

Electronic Warfare (EW) receivers are passive systems that are designed to detect and identify active radar emitters in the environment. The radar pulses emitted by multiple sources are received and must be deinterleaved, in other words, sorted according to the waveform to which they belong, i.e. according to their emitter. As radar signals are more complex and observations are denser, new Machine Learning (ML) approaches appear in the literature to enhance traditional Radar Signal Deinterleaving (RSD). In this paper, we propose an overview of the ML approaches to RSD. To this end, we identify some criteria to characterize a method: its used technique, underlying assumptions, exploited parameters, input characterization, and architectural pattern. First, the problem of RSD is detailed with its challenges and operational requirements. We then outline the methods of the literature inside a taxonomy based on the technique criterion : the first category includes PRI estimation methods including histogram-based methods, and ML techniques constitute three other categories : clustering, RNN, CNN. Finally, the other identified criteria are explained and discussed.

This paper presents a performance evaluation of stock picking by merging several technical indicators. Several fusion operators have been proposed either in the probabilistic or in the possibilistic framework. The latter fuzzy framework has been introduced to manage the uncertain information embedded in financial time series due to human biases as studied by behavioral finance. Performances of portfolio resulting from the proposed systems are evaluated according to cumulative returns but also through a risk analysis point of view (Sharpe ratio). Two fusion mechanisms (one probabilistic, one possibilistic) aiming at discriminating common information from merged technical indicators, produce the higher portfolio performances. It also appears that selecting specific technical indicators affects the overall performances of the proposed stock picking systems. Indeed, studying the technical indicators selection through a shared/nonshared information point of view reveals that possibilistic framework is more robust to redundant sources than probabilistic framework. Effects of some parameters used in the fusion algorithms (amount of assets, window length analysis, ...) are also investigated. Results from all these tests clearly show the high potentiality of technical indicators fusion to improve portfolio performances. However, these first promising results have to be further inspected within wider contexts, as discussed at the end of the paper.

Automatic extraction of buildings in urban and residential scenes has become a subject of growing interest in the domain of photogrammetry and remote sensing, particularly since the mid-1990s. Active contour model, colloquially known as snake model, has been studied to extract buildings from aerial and satellite imagery. However, this task is still very challenging due to the complexity of building size, shape, and its surrounding environment. This complexity leads to a major obstacle for carrying out a reliable large-scale building extraction, since the involved prior information and assumptions on building such as shape, size, and color cannot be generalized over large areas. This paper presents an efficient snake model to overcome such a challenge, called Super-Resolution-based Snake Model (SRSM). The SRSM operates on high-resolution Light Detection and Ranging (LiDAR)-based elevation images—called z-images—generated by a super-resolution process applied to LiDAR data. The involved balloon force model is also improved to shrink or inflate adaptively, instead of inflating continuously. This method is applicable for a large scale such as city scale and even larger, while having a high level of automation and not requiring any prior knowledge nor training data from the urban scenes (hence unsupervised). It achieves high overall accuracy when tested on various datasets. For instance, the proposed SRSM yields an average area-based Quality of 86.57% and object-based Quality of 81.60% on the ISPRS Vaihingen benchmark datasets. Compared to other methods using this benchmark dataset, this level of accuracy is highly desirable even for a supervised method. Similarly desirable outcomes are obtained when carrying out the proposed SRSM on the whole City of Quebec (total area of 656 km²), yielding an area-based Quality of 62.37% and an object-based Quality of 63.21%.

In this paper, we analyze the levels of performance of the (full or reduced versions of) space-time adaptive processing (STAP) algorithms applied to high-frequency surface wave radars (HFSWRs) for marine target detection to verify whether this family outperforms the usual moving target indication (MTI) algorithms. In fact, STAP algorithms have well-known drawbacks (e.g., large amount of secondary data). The propagation and backscattering of radar signals are included in the covariance matrix calculation. Since the propagation and backscattering have to be derived from physical features of the sea (e.g., sea Doppler) and the atmosphere, we also include the difference induced by the waveform (pulsed radar/chirp) in our study. Moreover, since the inverse covariance matrix is used in the target detection algorithms, matrix conditioning is also inspected for practical implementation reasons. We numerically detail the performance (signal noise clutter ratio and conditioning number of the STAP algorithms) with regards to the MTI performance for several configurations of radar and sea state. These results show that the improvements of signal noise clutter ratio are counterbalanced by a raise of the conditioning number, except in the case of eigencanceler approach. The sea-state parameters have few effects on the performance unlike to radar parameters.

Applications based on synergistic integration of optical imagery and LiDAR data are receiving a growing interest from the remote sensing community. However, a misaligned integration of these datasets fails to fully profit from the potential of both sensors. An optimum fusion of optical imagery and LiDAR data requires an accurate registration. This is a complex problem since a versatile solution is still missing, especially when data are collected at different times, from different platforms, under different acquisition configurations. This paper presents a coarse-to-fine registration method of optical imagery with airborne LiDAR data acquired in such context. Firstly, a coarse registration involves processes of extraction and matching of building candidates from the two datasets. Then, a Mutual Information-based fine registration is carried out. It involves a super-resolution approach applied to LiDAR data to generate images with the same resolution as the optical image, and a local approach of transformation model estimation. The proposed method succeeds at overcoming the challenges associated with this difficult context. For instance, considering the experimented airborne LiDAR (2011) and orthorectified aerial imagery (2016) datasets, their spatial shift is reduced by 48.15% after the proposed coarse registration. Moreover, the incompatibility of size and spatial resolution is well addressed by the super-resolution. Finally, a high accuracy of dataset alignment is also achieved, highlighted by a 40-cm error based on a check-point assessment and a 64-cm error based on a check-pair-line assessment. These promising results enable further researches for a complete fusion methodology between these datasets in this challenging context.