Paula Gómez-Pérez

Retrieving Vegetation Reradiation Patterns by Means of Artificial Neural Networks

Modeling vegetation usually implies obtaining experimental data by means of measurement campaigns. In general, the most accurate models need the reradiation pattern of the vegetation volume under study, or some of its related parameters. Obtaining this function might not be easy, and the measurement procedure may introduce errors depending on the location of the vegetation mass. An accurate tool to simplify this process is presented, based on the ability of artificial neural networks to infer behavioral patterns. This proposal reduces the acquisition of the experimental data to a scarce set of angles around the tree or bush, which will then be used to train a neural network capable of retrieving the desired reradiation function desired.

Reduction of radar performance for target detection within forests

Abstract. Radar performance has improved substantially from the first prototypes until current modern systems. Nowadays, technical advances run parallel to the development of new materials and shapes able to shield different targets or even make them invisible to the radar. This paper introduces the shielding phenomena from another point of view, dealing with the usage of forested radio propagation channels to degrade radar performance. Based on measurements carried out in five different forest and meadow scenarios, we study the probability of detection of a target located within such environments. Depending on the type of target, the frequency, and the vegetated surroundings, we demonstrate that a forest can provide radar invisibility even to large targets, reducing the probability of detection to values below 0.1.

Modeling vegetation attenuation patterns: A comparison between polynomial regressions and artificial neural networks

Polynomial regressions have been widely used for modeling vegetation patterns. However, artificial neural networks provide more efficient, accurate and generalizable models than polynomial regressions. This paper compares both machine-learning techniques in terms of RMS error and training set size in order to demonstrate the superiority of neural networks over well-known methods as polynomial regressions.

Modeling indoor vegetation re-radiation pattern with dynamic multivariate polynomial regressions

Modeling vegetation re-radiation patterns has been a recurrent problem for industry and radio planners. Nowadays, several reliable models exist, but they are very complex to implement and need extensive measurement campaigns to validate their results, or they are too simplistic. The proposal of this contribution is to introduce the use of dynamic multivariate polynomial regressions to fit vegetation re-radiation patterns accurately in a simple form. A novel application is presented, which is able to adapt the complexity of the model obtained to any vegetation specie or frequency under study.

Quantifying the degradation on radar detection range through vegetation clutter

Surface radar systems are commonly used for urveillance purposes, both civil and military. Such systems have deal with the surrounding environment, which commonly cludes forested areas that might degrade the received signal, educing the radar coverage. This paper enunciates a way to valuate these losses, and provides its quantification for five ifferent forested scenarios.

Analysis of Spatial Diversity in Both Transmitter and Receiver on Low-Height Links at Forests

This work focuses on radio wave propagation within forested environments, at 5.8 GHz. Concretely, we explore the advantages of implementing spatial diversity in reception or even in both ends for improving the strength of the received signal in such environments, which could be useful in applications such as vehicle-to-infrastructure, vehicle-to-vehicle, or emergency communications. Measurements gathered at both evergreen and deciduous forests sustain the thesis. Once processed, the results support the proposal of implementing a spatial diversity technique in reception or in both ends using a 2 × 4 (or 2 × 2) scheme in order to improve the connectivity at 5.8 GHz band within forests. In fact, we estimated a gain due to spatial diversity in reception of 5 dB and 2 dB at evergreen and deciduous forests, respectively, and 16 dB or 5 dB when implementing at both ends.