Open-Loop Sidescan Sonar Mosaic and ANN Velocity Estimation

摘要

This paper presents a system to generate open-loop mosaics from the images obtained by a sidescan sonar using dead reckoning as a positioning system in a low cost ROV, which does not include a Doppler velocity log (DVL). The ROV velocity is then estimated using an artificial neural network (ANN) for feeding the dead reckoning positioning. The training process of the neural network is also described. The sidescan sonar readings are used, based on the currently estimated position, to update the sonar mosaic. The process of the transformations and corrections of the sidescan sonar are also described. Finally, some results obtained in a thermal power plant are presented.