| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 11.9 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
O posicionamento e navegação de veículos subaquáticos é um problema complexo. Dada
a hostilidade do meio aquático, os métodos de localização existentes (geralmente baseados
em acústica) são limitados e pouco precisos, particularmente em operações junto a infraestruturas
subaquáticas como oleodutos no fundo marinho profundo. Consequentemente para
determinar a posição de veículos subaquáticos é necessário recorrer, entre outras, a técnicas
alternativas como técnicas de visão artificial. Neste trabalho propõe-se uma solução que
converge para a solução dos problemas referidos, baseada em visão stereo.
Foram estudadas técnicas de visão adequadas para localização de veículos subaquáticos,
com auxílio das bibliotecas Open Source do OpenCV [1], desenvolvido um sistema de
posição relativa baseada em visão e obtidos resultados experimentais em cenários de operação
com um pequeno ROV (nomeadamente num tanque de testes e em exercícios navais
REP2012). Finalmente, foi desenvolvido um sistema de localização externa com visão stereo
para o tanque de testes do LSA, tendo o mesmo sido validado e caracterizado experimentalmente.
Os resultados obtidos demonstram que o sistema proposto é uma solução viável, e até
bastante fiável, para a detecção e localização relativa ou absoluta de veículos subaquáticos,
visto que para todos os métodos de visão implementados, o erro obtido foi consideravelmente
baixo para as aplicações em questão. O sistema desenvolvido foi ainda comparado com
um equipamento de localização de sinais acústicos (USBL), num cenário que evidencia as
limitações dos sistemas de posicionamento acústico (tanque de testes) e comprovou-se que a
solução proposta obteve melhores resultados. Deste modo, o sistema implementado constitui
uma alternativa funcional aos seus equivalentes acústicos.
The navigation and positioning of underwater vehicles is a complex problem to solve, due to the harshness of the surrounding environment and the limitations and low precision of the existent localization methods (usually based on acoustic technology). These are particularly noticeable in repair operations of underwater structures, oil-pipe maintenance and mapping of the seabed. For this reason, the determination of the position of underwater vehicles alternatively recurs to artificial vision techniques, to solve the problem of the positioning relative to an underwater structure or even the absolute position. In the present work a solution for both of these problems, based on stereo vision, is proposed. Several vision techniques, with resource to the Open Source OpenCV [1] libraries, suitable for underwater vehicles localization were studied. Furthermore, a vision based relative position system was developed and experimental results were obtained in operational scenarios with a small ROV (namely in a test water tank and in the REP2012 naval drills under the Portuguese navy). Finally, it was developed a stereo vision external localization system for the LSA test water tank. This system was then experimentally validated and fully characterized. The obtained results show that the proposed system is a viable solution, and quite reliable, for detection and relative or absolute localization of underwater vehicles, once for all the implemented vision methods the obtained error was considerably low for the applications in cause. The developed system was also compared with an acoustic localization sensor (USBL), in a scenario that points out the limitations of these type of sensors (water tank). It was proven that the proposed solution has better results. Thus, the implemented system is a functional alternative to its acoustic equivalents.
The navigation and positioning of underwater vehicles is a complex problem to solve, due to the harshness of the surrounding environment and the limitations and low precision of the existent localization methods (usually based on acoustic technology). These are particularly noticeable in repair operations of underwater structures, oil-pipe maintenance and mapping of the seabed. For this reason, the determination of the position of underwater vehicles alternatively recurs to artificial vision techniques, to solve the problem of the positioning relative to an underwater structure or even the absolute position. In the present work a solution for both of these problems, based on stereo vision, is proposed. Several vision techniques, with resource to the Open Source OpenCV [1] libraries, suitable for underwater vehicles localization were studied. Furthermore, a vision based relative position system was developed and experimental results were obtained in operational scenarios with a small ROV (namely in a test water tank and in the REP2012 naval drills under the Portuguese navy). Finally, it was developed a stereo vision external localization system for the LSA test water tank. This system was then experimentally validated and fully characterized. The obtained results show that the proposed system is a viable solution, and quite reliable, for detection and relative or absolute localization of underwater vehicles, once for all the implemented vision methods the obtained error was considerably low for the applications in cause. The developed system was also compared with an acoustic localization sensor (USBL), in a scenario that points out the limitations of these type of sensors (water tank). It was proven that the proposed solution has better results. Thus, the implemented system is a functional alternative to its acoustic equivalents.
Description
Mestrado em Engenharia Electrotécnica e de Computadores.Área de Especialização de Sistemas Autónomos
Keywords
Groudtruth Stereo Subaquático Posicionamento Detecção Infraestruturas subaquáticas Visão subaquática Veículos subaquáticos Tracking
Citation
Publisher
Instituto Politécnico do Porto. Instituto Superior de Engenharia do Porto