@InProceedings{FornariSantShig:2018:SeApAu,
author = "Fornari, Gabriel and Santiago J{\'u}nior, Valdivino Alexandre de
and Shiguemori, Elcio",
affiliation = "{Instituto Nacional de Pesquisas Espaciais (INPE)} and {Instituto
Nacional de Pesquisas Espaciais (INPE)}",
title = "A self-adaptive approach for autonomous UAV navigation via
computer vision",
year = "2018",
organization = "International Conference on Computational Science and its
Applications, 18. (ICCSA)",
keywords = "Unmanned Aerial Vehicles, Computer vision Autonomous navigation,
Self-adaptive.",
abstract = "In autonomous Unmanned Aerial Vehicles (UAVs), the vehicle should
be able to manage itself without the control of a human. In these
cases, it is crucial to have a safe and accurate method for
estimating the position of the vehicle. Although GPS is commonly
employed in this task, it is susceptible to failures by different
means, such as when a GPS signal is blocked by the environment or
by malicious attacks. Aiming to fill this gap, new alternative
methodologies are arising such as the ones based on computer
vision. This work aims to contribute to the process of autonomous
navigation of UAVs using computer vision. Thus, it is presented a
self-adaptive approach for position estimation able to change its
own configuration for increasing its performance. Results show
that an Artificial Neural Network (ANN) presented the best
performance as an edge detector for pictures with buildings or
roads and the Canny extractor was better at smooth surfaces.
Moreover, our selfadaptive approach as a whole shows gain up to
15% if compared with non-adaptive methodologies.",
conference-location = "Melbourne, Australia",
conference-year = "02-05 July",
language = "en",
urlaccessdate = "29 mar. 2024"
}