Abstract

Abstract

Abstract
The rising needs for increased energy efficiency and better use of renewable energy sources bring out the necessity for improved energy management and forecasting models. Electricity consumption, in particular, is subject to large variations due to the effect of multiple variables, such as the temperature, luminosity or humidity, and of course, consumers' habits. Current forecasting models are not able to deal adequately with the influence and correlation between the multiple involved variables. Hence, novel, adaptive forecasting models are needed. This paper presents a novel approach based on multiple artificial intelligence-based forecasting algorithms. The considered algorithms are artificial neural networks, support vector machines hybrid fuzzy inference systems, Wang and Mendel's fuzzy rule learning method and a genetic fuzzy system for fuzzy rule learning based on the MOGUL methodology. These algorithms are used to forecast the electricity consumption of a real office building, using multiple input variables and consumption disaggregation. © 2019 IEEE.

Abstract
Virtual Reality (VR) has been recently gaining interest from researchers and companies, contributing to the development of the associated technologies that aim to transport its users to a virtual environment by the stimulation of their senses. Technologies such as Head-Mounted Displays (HMD), capable of presenting 360 degrees video in 3D, are becoming affordable and, consequently, more common among the average consumer, potentiating the creation of a market for VR experiences. The purpose of this study is to measure the influence of (a) video format (2D/monoscopic vs 3D/stereoscopic), (b) sound format (2D/stereo vs 3D/spatialized), and (c) gender on users' sense of presence and cybersickness, while experiencing a VR application using an HMD. Presence and cybersickness were measured using questionnaires as subjective measures. Portuguese versions of the Igroup Presence Questionnaire for presence and the Simulator Sickness Questionnaire for cybersickness were used. Results revealed no statistically significant differences between (a) VIDEO and (b) SOUND variables on both senses of presence and cybersickness. When paired with (a) VIDEO, the independent variable (c) Gender showed significant differences on almost all subscales of presence. Results suggest that the widely acknowledged differences in spatial ability between genders were a major factor contributing to this outcome.

Abstract
This paper explores the usage of unmanned aerial vehicles (UAVs) to acquire remotely sensed very high-resolution imagery for classification of an agrosilvopastoral system in a rural region of Portugal. Aerial data was obtained using a low-cost UAV, equipped with an RGB sensor. Acquired imagery undergone a photogrammetric processing pipeline to obtain different data products: an orthophoto mosaic, a canopy height model (CHM) and vegetation indices (VIs). A superpixel algorithm was then applied to the orthophoto mosaic, dividing the images into different objects. From each object, different features were obtained based in its maximum, mean, minimum and standard deviation. These features were extracted from the different data products: CHM, VIs, and color bands. Classification process – using random forest algorithm – classified objects into five different classes: trees, low vegetation, shrubland, bare soil and infrastructures. Feature importance obtained from the training model showed that CHM-driven features have more importance when comparing to those obtained from VIs or color bands. An overall classification accuracy of 86.4% was obtained.

Abstract