Localização topológica baseada em visão computacional para aplicação em veículos inteligentes
Authors
Antonio Rola Neto
Laboratório de Robótica, sistemas inteligentes e Complexos - RobSIC Instituto de Ciências Tecnológicas Universidade Federal de Itajubá, Campus Itabira, MG.
Giovani Bernardes Vitor
Laboratório de Robótica, sistemas inteligentes e Complexos - RobSIC Instituto de Ciências Tecnológicas Universidade Federal de Itajubá, Campus Itabira, MG.
Tiago Gaiba de Oliveira
Laboratório de Robótica, sistemas inteligentes e Complexos - RobSIC Instituto de Ciências Tecnológicas Universidade Federal de Itajubá, Campus Itabira, MG.
In the field of intelligent vehicles, two essential items are mapping and localization. Therefore, this work proposes a case study that uses topological localization, a feasible alternative to address unresolved challenges within the specific context of localization in outdoor environments and urban centers. For this purpose, a computer vision strategy is employed to identify street intersections in the process of topological localization and navigation. Thus, the Speeded Up Robust Features (SURF) algorithm will be used to recognize key points in images along the movement of the camera attached to the vehicle. These key points are used for image association and identification via the K-Nearest Neighbors Match (KNN Match), and finally, the RANdom SAmple Consensus (RANSAC) is applied to perform outlier filtering and achieve more robust identification. The qualitative and quantitative results demonstrate the developed system as well as an identification capability that reached 84.4% in challenging scenarios.
