Captura e análise da percepção de cenas urbanas em diferentes escalas.

Abstract

Urban perception concept can be seen as the understanding, the meaning that people give to their urban spaces. Such perception are of great importance since people guide their actions in space according to how they perceive the spaces around them. For this reason, several studies have been performed over the years seeking to better understand this relationship between people and urban spaces in order to produce knowledge that can be used to improve urban projects and space management. Traditional methods used by such studies involved site visits, interviews and the use of questionnaires, which could present or not images of the spaces being evaluated. Computing advancement has brought solutions and methods that can contribute to greater agility in data capture, reducing costs and increasing the number of participants in the research. Crowdsourcing is one of these solutions. In this context, this thesis investigated the use of crowdsourcing to conduct urban perception studies focusing both on a larger scale, considering urban scenes from several points of the city, as well as on a smaller scale, considering a larger number of scenes for specific sites in the city. Two major steps of urban scenes and perception collection were performed, focusing on the city of Campina Grande, Paraíba. In a first study, which focus on a larger scale, we evaluated a method to relate gathered perceptions, urban space characteristics and sociodemographic profile of participants. The results showed, for example, that participants evaluated the scenes with more trees and better maintenance condition of their elements as more pleasant, and the scenes that had better maintenance condition of their elements and more people as safer. In addition, it was possible to find some perception differences between groups of participants and relate these differences with urban space characteristics. For example, men and high income people have preferred even more well-maintained places than women and low income people. In a second study, which focus on a smaller scale, we evaluated the use of a method that calculates Bayesian surprises to find urban scenes that stand out from their neighborhood. The results showed that among scenes highlighted by the method it was possible to find scenes with technical virtues and problems, pointed out by urbanists, as well as it was possible to recognize relevant and actionable scenes, towards management teams, from the point of view of research participants and street users.