Abstract
In recent years we witnessed the arrival of new trends, such as server virtualization and cloud services, an increasing number of mobile devices and online contents, leading the networking industry to deliberate about how traditional network architectures can be adapted or even deciding if a new perspective for them should be taken. SDN (Software-Defined Networking) emerged under this framing, opening a road for new developments due to the centralized logic control and view of the network, the decoupling of data and control planes, and the abstraction of the underlying network infrastructure from the applications. Although firstly oriented to packet switching, network measurements have also emerged as one promising field for SDN, as its flexibility enables programmable measurements, allowing a SDN controller to manage measurement tasks concurrently at multiple spatial and temporal scales. In this context, this paper is focused on exploring the SDN architecture and components for supporting the flexible selection and configuration of network monitoring tasks that rely on the use of traffic sampling. The aim is to take advantage of the integrated view of SDN controllers to apply and configure appropriate sampling techniques in network measurement points according to the requirements of specific measurement tasks. Through SDN, flexible and service-oriented configuration of network monitoring can be achieved, allowing also to improve the trade-off between accuracy and overhead of the monitoring process. In this way, this study, examining relevant SDN elements and solutions for deploying this monitoring paradigm, provides useful insights to enhance the programmability and efficiency of sampling-based network monitoring. © 2017, Springer International Publishing AG.

Abstract
The problem of Label Ranking is receiving increasing attention from several research communities. The algorithms that have been developed/adapted to treat rankings of a fixed set of labels as the target object, including several different types of decision trees (DT). One DT-based algorithm, which has been very successful in other tasks but which has not been adapted for label ranking is the Random Forests (RF) algorithm. RFs are an ensemble learning method that combines different trees obtained using different randomization techniques. In this work, we propose an ensemble of decision trees for Label Ranking, based on Random Forests, which we refer to as Label Ranking Forests (LRF). Two different algorithms that learn DT for label ranking are used to obtain the trees. We then compare and discuss the results of LRF with standalone decision tree approaches. The results indicate that the method is highly competitive.

Abstract
The construction of earthen embankment terraces in the Douro Region raises a set of problems related to hydrological processes. The main objective of this study is the evaluation of the spatial variation of electrical resistivity in agriculture terraces at Douro valley (Portugal). To achieve this objective, two variables are analysed, the soil electrical resistivity and the flow direction algorithm. In a field survey we recorded 13 electrical resistivity profiles. The contributing area was calculated with the algorithms D infinity (Deterministic Infinity Flow) and MFD (Multiple Flow Direction) and the results are the base of the internal runoff modelling, both supported by the digital elevation model with a spatial resolution of 1m2. A correlation between the spatial variation of the soil electrical resistivity represented by the standard deviation of the electrical resistivity for each profile and the average value of the contributing area coincident with each profile was established. The electrical resistivity standard deviation seems to be moderately well correlated according to the D infinity algorithm at about 1m of depth, and it has a good correlation at 1,5m to 2m of depth with the MFD algorithm. Taken together, the results show a significant positive statistical correlation between the electrical resistivity standard deviation and the contributing areas (MFD and D infinity) depending on the soil depth.

Abstract
Object localization and tracking is core to many practical applications, like human-computer interaction, security and surveillance, robot competitions and Industry 4.0. Such task may be computationally demanding, especially for traditional embedded systems, that usually have tight processing and storage constraints. This calls for the investigation of alternatives, including emergent heterogeneous embedded systems, like the Parallella line of single-board-computers (SBCs). The work presented in this paper explores the use of a Parallella board with a 16-core Epiphany co-processor, to perform real-time tracking of objects in frames captured by a Kinect sensor, based on color segmentation. We addressed several processing strategies, trying to assess which one performed better. We also ran the same code (where applicable) in several models of the Raspberry Pi platform, for comparison. We conclude that effectively exploring the Epiphany co-processor is not trivial, requiring considerable programming effort and suitable applications (CPU-demanding and highly parallelizable), to the extent that simpler development approaches, on more recent SBCs may be more effective. © 2017.

Abstract
The usage of small-sized unmanned aerial systems (UAS) has increased in the last years, in many different areas, being agriculture and forestry those who benefit the most from this relatively new remote sensing platform. Leaf area index, canopy and plant volume are among the parameters that can be determined using the very high resolution aerial data obtained by sensors coupled in unmanned aerial vehicles (UAV). This remote sensing technology affords the possibility of monitoring the vegetative development, identifying different types of issues, enabling the application of the most appropriated treatments in the affected areas. In this paper, a methodology allowing to perform multi-temporal UAS-based data analysis obtained by different sensors is proposed. A case study in vineyards and chestnuts is used to prove the benefits of continuous crop monitoring in its management and productivity of agroforestry parcels/activities. (C) 2017 The Authors. Published by Elsevier B.V.

Abstract
The autonomy of robotic underwater vehicles is dependent on the ability to perform long-term and long-range missions without need of human intervention. While current state-of-the-art underwater navigation techniques are able to provide sufficient levels of precision in positioning, they require the use of support vessels or acoustic beacons. This can pose limitations on the size of the survey area, but also on the whole cost of the operations. Terrain Based Navigation is a sensor-based navigation technique that bounds the error growth of dead reckoning using a map with terrain information, provided that there is enough terrain variability. An obvious advantage of Terrain Based Navigation is the fact that no external aiding signals or devices are required. Because of this unique feature, terrain navigation has the potential to dramatically improve the autonomy of Autonomous Underwater Vehicles (AUVs). This paper consists on a comprehensive survey on the recent developments for Terrain Based Navigation methods proposed for AUVs. The survey includes a brief introduction to the original Terrain Based Navigation formulations, as well as a description of the algorithms, and a list of the different implementation alternatives found in the literature. Additionally, and due to the relevance, Bathymetric SLAM techniques will also be discussed.